CN102421916A

CN102421916A - Method for evaluating cancer

Info

Publication number: CN102421916A
Application number: CN2010800179242A
Authority: CN
Inventors: 黑田雅彦; 斋藤彰; 佐野真贵
Original assignee: NEC Corp; Tokyo Medical University
Current assignee: NEC Corp; Tokyo Medical University
Priority date: 2009-04-21
Filing date: 2010-04-21
Publication date: 2012-04-18
Also published as: US20120115139A1; US20140178874A1; EP2423323A4; EP2423323A1; WO2010123043A1; JPWO2010123043A1

Abstract

Disclosed is a method for evaluating cancer by using a novel cancer marker for evaluating cancer onset, preclinical stage, clinical stage or prognosis in a subject. A novel cancer marker for evaluating cancer which comprises at least one miRNA selected from hsa-miR-92 and hsa-miR-494. The aforesaid cancer marker in a cell or tissue sample is detected and cancer risk in the sample is evaluated based on the expression level thereof. According to this evaluation method, the cancer risk in the sample can be evaluated at a high reliability by detecting the aforesaid miRNA that is a cancer marker. Concerning the method for detecting the cancer marker as described above, it is preferred to carry out, for example, in situ hybridization on the immobilized sample with the use of a labeled probe.

Description

The method that is used for assessment of cancer

Technical field

The present invention relates to come the method for the possibility of assessment of cancer through detecting new cancer markers.

Background technology

In the field of clinical medicine practise, require easily to confirm the existence of disease or do not exist, the progress degree of disease, the effect that after treatment, obtains etc.In these cases, the method as being used for indirectly confirming them has proposed the detection mark, and the expression amount of this mark is followed outbreak or the progress of every kind of disease and specificity changes, and has in fact carried out it is dropped into the actual trial of using.

In various diseases, early detection is as the malignant tumour of so-called cancer and suitably select and change the therapeutic strategy particularly important to it.Therefore, in order to realize the aforesaid indirect judgement of passing through the detection of mark, various cancer markers have been reported.Cancer markers also is called as tumor markers.The concrete example of cancer markers comprises PSA (PSA), CEA (CEACAMS), CA 19-9 (CEACAMS 19-9) and CA 72-4 (CEACAMS 72-4).And, the down-regulated expression (non-patent document 1 and 2) of miRNA such as has-mir-15, has-mir-16, miR-143 and miR-145 in Lymphocytic leukemia and colorectal carcinoma etc. has been described in non-patent document 1 and 2.

[reference listing]

[non-patent document]

[non-patent document 1] Calin GA; Dumitru CD, Shimizu M et al., Frequent deletions and down-regulation of micro-RNA genes miR15 and miR16 at 13q14 in chronic lymphocytic leukemia; Proc Natl Acad Sci USA; 2002, vol.99, pp.15524-9

[non-patent document 2] Michael MZ, SM OC, van Holst Pellekaan NG, Young GP; James RJ, Reduced accumulation of specific microRNAs in colorectal neoplasia, Mol Cancer Res; 2003, vol.1, pp.882-91

Summary of the invention

The problem that the present invention will solve

Yet, in the field of clinical medicine practise, need cancer markers, use cancer markers good reliability to confirm the outbreak of cancer and their progress.Therefore, still need providing of new cancer markers.Therefore, for aforementioned idea, the purpose of this invention is to provide a kind of new cancer markers appraisal procedure of coming assessment of cancer and assessment reagent that will in this appraisal procedure, use of using.

The means that are used to deal with problems

The invention provides a kind of appraisal procedure that is used for the possibility of assessment of cancer, may further comprise the steps: detect the cancer markers in sample; And, be evaluated at the said cancer possibility in the said sample based on the expression level of said cancer markers.In this appraisal procedure, said sample is a cell or tissue, and said cancer markers comprises at least one miRNA that is selected from hsa-miR-92 and hsa-miR-494.

Effect of the present invention

Contriver of the present invention has carried out diligent research, and as a result of, and they find that hsa-miR-92 and the hsa-miR-494 expression level in cell or tissue follows the development of cancer and change, and realize the present invention thus.According to appraisal procedure of the present invention,, might for example confirm existing or not existing of cancer development or cancer progression with good reliability through detecting at least one the expression level of above-mentioned miRNA in sample.And, for example, see the marked difference in the expression of these miRNA between to the canceration feminine gender and the positive.Therefore,, become and easily to detect cancer in the early stage, and such detection is difficult through general palpation etc. according to the present invention.And, have the more cancer assessment of the high reliability possibility that becomes through appraisal procedure of the present invention being associated with the cancer assessment of for example using traditional HE dyeing etc., making.

Description of drawings

Fig. 1 is the block diagram that illustrates according to the cancer pathology image diagnosis support system of embodiment of the present invention 1B.

Fig. 2 is shown in the content of the dyeing image data base among embodiment of the present invention 1B.

Fig. 3 is shown in the schema according to an example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1B shown in Fig. 1.

Fig. 4 is shown in the schema according to another example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1B shown in Fig. 1.

Fig. 5 is shown in the schema according to another example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1B shown in Fig. 1.

Fig. 6 is shown in the schema according to another example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1B shown in Fig. 1.

Fig. 7 is the block diagram that illustrates according to the cancer pathology image diagnosis support system of embodiment of the present invention 1C.

Fig. 8 is shown in the schema according to an example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1C shown in Fig. 7.

Fig. 9 is shown in the schema according to another example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1C shown in Fig. 7.

Figure 10 is shown in the schema according to another example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1C shown in Fig. 7.

Figure 11 is the block diagram that illustrates according to the cancer pathology image diagnosis support system of embodiment of the present invention 1D.

Figure 12 is the block diagram that illustrates according to the cancer pathology image diagnosis support system of embodiment of the present invention 1E.

Figure 13 is the block diagram that illustrates according to the configuration of the system of the execution feature selection approach of embodiment of the present invention 2A.

Figure 14 is the schema of diagram according to the feature selection approach of embodiment of the present invention 2B.

Figure 15 is the schema of example of processing that is used to set up the classification table that illustrates according to embodiment of the present invention 2B.

Figure 16 is illustrated in the example of the classification table of setting up among embodiment of the present invention 2B.

Figure 17 is the block diagram that is illustrated in the configuration of the system that is used to diagnose among embodiment of the present invention 2B.

Figure 18 is the schema that is illustrated in the example of the processing that is used to extract subimage among embodiment of the present invention 2B.

Figure 19 is the block diagram that illustrates according to the configuration of the cancer pathology image diagnosis support system of embodiment of the present invention 1A.

Figure 20 is the block diagram that illustrates according to the cancer pathology image diagnosis support system of embodiment of the present invention 1A.

Figure 21 is shown in the schema according to an example of the operation of the cancer pathology image diagnosis support system of embodiment of the present invention 1A shown in Figure 20.

Figure 22 illustrates the photo that is used for being illustrated in embodiments of the invention 1 leukocytic miRNA coloration result.

Figure 23 illustrates the photo of the miRNA coloration result that is used for being illustrated in embodiments of the invention 2 breast tissues.

Figure 24 illustrates the photo that is used for being illustrated in embodiments of the invention 3 hepatocellular miRNA coloration results.

Embodiment

The implication of the corresponding term that uses in the present invention is following.Malignant tumour generally represented in term " cancer ".The outbreak of term " canceration " ordinary representation cancer, and also comprise " vicious transformation ".About term " outbreak ", for example, a people is diagnosed as the outbreak that the time point with specific disorders is called as disease through the comprehensive judgement based on disease specific clinical symptom or analysis data etc.Term " preclinical phase " is commonly referred to as the situation before seizure of disease, and wherein, the disease specific clinical symptom does not also occur, but is in the early stage of disease, and wherein, the malignant cell of trace exists.Term " prognosis " expression is for example such as the course of disease after the disease treatment of operation.Because the cancer markers of using in the present invention can be provided for the for example useful information of predict prognosis, the prediction course of disease and selection suitable therapeutic method, so it also can be called as " prognosis factor ".Can wait suitably definite " cancer progression phase " based on the kind of for example cancerous tissue.Usually, 0 phase and I phase can be classified as the initial stage cancer, and the II phase can be classified as early-stage cancer, and III phase and IV phase can be classified as TCA.

In the present invention, " possibility of cancer " for example comprises the possibility that the person under inspection possibly suffer from cancer, no matter and canceration, cancer progression phase such as preclinical phase or clinical phase or prognosis state etc. whether have occurred.

< cancer markers >

Cancer markers miRNA in the present invention is at least one miRNA that is selected from hsa-miR-92 and hsa-miR-494 as stated.Below, cancer markers also is called as " cancer markers miRNA ".

In the present invention, cancer markers miRNA can for example be strand (monomer) or double-stranded (dimer).And in the present invention, cancer markers miRNA can for example be jejune miRNA or sophisticated miRNA.The example of immature miRNA comprises initial miRNA of primary transcript (pri-miRNA) and precursor miRNA (pre-miRNA).Pri-miRNA has because of the hairpin ring structure of intramolecularly due to combining.Use Drosha that pri-miRNA is cut, convert pri-miRNA into short pre-miRNA thus with stem-ring structure.Below, pre-miRNA also is called as stem-ring miRNA.Use Dicer that pre-miRNA is cut, thus, produce shorter double-stranded RNA (miRNA-miRNA ^*).This double-stranded RNA is untied on RISC, produces two single stranded RNAs thus.Single stranded RNA is ripe miRNA.Below, one of ripe miRNA is called as function miRNA, and another is called as small miRNA ^*

In the present invention, cancer markers miRNA is by special restriction, and preferably stem-ring miRNA or ripe miRNA, especially preferably ripe miRNA.

The example of hsa-miR-92 comprises hsa-miR-92a and hsa-miR-92b.

Hsa-miR-92a can be immature miRNA or ripe miRNA, as stated.Below, the former also is called as immature hsa-miR-92a, and the latter also is called as ripe hsa-miR-92a.

The example of immature hsa-miR-92a comprises hsa-miR-92a-1 and hsa-miR-92a-2, and immature hsa-miR-92a can be they any one.Hsa-miR-92a-1 and hsa-miR-92a-2 are stem-ring miRNA.Below, the former also is called as stem-ring hsa-miR-92a-1, and the latter also is called as stem-ring hsa-miR-92a-2.They are the transcription products that draw from different genome areas, and still, the sequence of their mature form is identical.The sequence of stem-ring has-miR-92a-1 for example is registered under the accession number MI0000093, and the sequence of stem-ring has-miR-92a-2 for example is registered under the accession number MI0000094.

Ripe has-miR-92a can for example be function miRNA, and its sequence for example is registered under the accession number MIMAT0000092.The sequence of this functional has-miR-92a is illustrated among the following SEQ ID NO:1.

Functional has-miR-92a (SEQ ID NO:1)

5’-uauugcacuugucccggccugu-3’

Other examples of ripe has-miR-92a comprise small miRNA ^*Small miRNA ^*Example comprise has-miR-92a-1 ^*And has-miR-92a-2 ^*Has-miR-92a-1 ^*Sequence for example be registered under the accession number MIMAT0004507.This sequence is illustrated among the SEQ ID NO:6.Has-miR-92a-2 ^*Sequence be registered under the accession number MIMAT0004508.This sequence is illustrated among the SEQ ID NO:7.

Small hsa-miR-92a-1 ^*(SEQ ID NO:6)

5’-agguugggaucgguugcaaugcu-3’

Small hsa-miR-92a-2 ^*(SEQ ID NO:7)

5’-ggguggggauuuguugcauuac-3’

Though has-miR-92b is that the kind subsequence of has-miR-92b and has-miR-92a's is similar from the transcription product that obtains with has-miR-92a different gene group zone.Therefore, similar with has-miR-92a, has-miR-92b can be used as cancer markers.Has-miR-92b can be immature miRNA or initial miRNA, as stated.Below, the former also is called as immature has-miR-92b, and the latter also is called as ripe has-miR-92b.

In various immature has-miR-92b, stem-also be called as below ring miRNA stem-ring has-miR-92b.The sequence of stem-ring has-miR-92b for example is registered under the accession number MI0003560.

Ripe has-miR-92b can be for example functional miRNA, and its sequence is registered under the accession number MIMAT0003218.The sequence of this functional has-miR-92b is illustrated among the SEQ ID NO:3.

Functional has-miR-92b (SEQ ID NO:3)

5’-uauugcacucgucccggccucc-3’

Other examples of ripe has-miR-92b comprise small miRNA ^*Small miRNA ^*Example comprise has-miR-92b ^*Though has-miR-92b ^*Be from has-miR-92a-1 ^*Or has-miR-92a-2 ^*The transcription product that different gene group zone obtains, but its kind subsequence and has-miR-92a-1 ^*Or has-miR-92a-2 ^*Similar.Therefore, with hsa-miR-92a-1 ^*Or hsa-miR-92a-2 ^*Similar, has-miR-92b ^*Can be used as cancer markers.Has-miR-92b ^*Sequence for example be registered under the accession number MIMAT0004792.In this sequence shown in the SEQ ID NO:4.

Small hsa-miR-92b ^*(SEQ ID NO:4)

5’-agggacgggacgcggugcagug-3’

Has-miR-494 can be immature miRNA or ripe miRNA, as stated.Below, the former also is called as immature has-miR-494, and the latter also is called as ripe has-miR-494.

The example of immature has-miR-494 comprises stem-ring miRNA.Below, this also is called as stem-ring has-miR-494.The sequence of stem-ring has-miR-494 for example is registered under the accession number MI0003134.

The example of ripe has-miR-494 comprises function miRNA, and its sequence for example is registered under the accession number MIMAT0002816.The sequence of this functional has-miR-494 is illustrated among the SEQ ID NO:2..

Hsa-miR-494 under the function (SEQ ID NO:2)

5’-ugaaacauacacgggaaaccuc-3’

As disclosed in the file of listing below, for example, 5 ' end and the 3 ' end of each miRNA have some variations respectively.Therefore, each miRNA in the present invention also comprises version, and each version has the different of some bases with the sequence of its mature form.

Wu?H.et?al.，2007，PLoS?ONE?2(10)：e1020?miRNA?profiling?of?naive，effector?and?memory?CD8?T?cells

Pablo?Landgraf?et?al.，2007，Cell，vol.129，pp.1401-1414?A?Mammalian?microRNA?Expression?Atlas?Based?on?Small?RNA?Library?Sequencing

Neilson?et?al.，2007，Genes?Dev，vol.21，pp.578-589?Dynamic?regulation?of?miRNA?expression?in?order?to?stage?of?cellular?development

Ruby?et?al.，2006，Cell，vol.127，pp.1193-1207?Large-scale?sequencing?reveals?21U-RNAs?and?additional?microRNAs?and?endogeneous?siRNAs?in?C.elegans

Obernoster?et?al.，RNA?2006?12：pp.1161-1167?Post-transcriptional?regulation?of?microRNA?expression

Lagos-Quintana?et?al.，2002，Curr?Biol，vol.12，pp.735-739Identification?of?tissue-specific?microRNAs?from?mouse

Cancer markers miRNA in the present invention comprises, for example: have the polynucleotide with the base sequence homologous base sequence of corresponding sequence identifier; And, have polynucleotide with its complementary base sequence.Identity degree when " homology " refers to the sequence that will be compared to each other when suitable comparison between them, and be illustrated in the appearance ratio (%) that Nucleotide matees fully between these sequences.When the base sequence of the base sequence of having described polynucleotide and miRNA of the present invention " had homology ", its these polynucleotide of expression and this miRNA were similar as to be enough to keep the function as miRNA of the present invention.Comparison for example, can realize such as any particular algorithms of BLAST through using.Even when base sequence difference each other is for example point mutation such as replacement, disappearance or adds, we can say that also they are homologous, as long as such difference does not influence the function of miRNA.The quantity of different bases for example is 1 to 20,1 to 15,1 to 10,1 to 5,1 to 3,1 to 2 or 1 between base sequence.And, when the base sequence of two polynucleotide that will be compared to each other for example has 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99% or 100% identity, we can say that they are homologous.And, for example,, we can say that these two polynucleotide are homologous when one of two polynucleotide during with the multi-nucleotide hybrid that has under stringent condition with another polynucleotide complementary base sequence.This stringent condition is by special restriction, and can be for example two polynucleotide to be comprised 6 * SSC, 0.5%SDS, remains in the solution of the salmon sperm nucleic acid of 5 * Denhardt solution and 0.01% sex change under the temperature " Tm (℃)-25 ℃ " to spend the night.

< appraisal procedure >

As stated, appraisal procedure according to the present invention may further comprise the steps: detect the cancer markers in sample; And, be evaluated at the possibility of the cancer in the sample based on the expression level of said cancer markers.In this appraisal procedure, sample is a cell or tissue, and cancer markers comprises at least one miRNA that is selected from has-miR-92 and has-miR-494.According to the present invention; For example; Might assess the possibility that person under inspection for example possibly suffer from cancer through detecting the cancer markers in person under inspection's sample, becoming, no matter and canceration, cancer progression phase such as preclinical phase (initial stage) or clinical phase or prognosis state etc. whether have occurred.Appraisal procedure of the present invention also can for example be called as and a kind ofly be used to assess the person under inspection and whether have method for cancer.

Cancer markers of the present invention as stated.The cancer markers that detect in the present invention, for example can be that any one of has-miR-92 and has-miR-494 maybe can be they both.

When the cancer markers that will detect for example was has-miR-92, it can be that any one of ripe has-miR-92 and immature has-miR-92 maybe can be they both.The kind of ripe has-miR-92 and immature has-miR-92 is not limited yet, and, for example, can use their any, their two kinds or their all categories.

When the cancer markers that will detect was ripe has-miR-92, it can be, for example: functional has-miR-92, small has-miR-92a-1 ^*With small has-miR-92a-2 ^*Any one; Their any two; Or theirs is whole.When the cancer markers that will detect is immature has-miR-92a, it can for example be stem-ring has-miR-92a-1 and stem-ring has-miR-92a-2 any one or they both.When the cancer markers that will detect was ripe has-miR-92b, it can for example be functional has-miR-92b and small has-miR-92b ^*Any one or they both.

When the cancer markers that will detect was has-miR-494, it can for example be that any one of ripe has-miR-494 and immature has-miR-494 maybe can be they both.The kind of ripe has-miR-494 and immature cancer markers miRNA is not limited yet, and, for example, can use their any, their two kinds or their all categories.Under the situation of ripe has-miR-494, its example comprises functional has-miR-494.

The invention is characterized in the expression level that detects cancer markers, as stated.The method that is used to detect cancer markers is never limited, and, can use currently known methods.This detection method preference comprises the visualization method of cancer markers in this way.Be used to realize visualization method not by special restriction, and preferably wait and realize that this is visual through for example colour developing, fluorescence or radioautograph.When coming visual cancer markers, can for example wait and detect cancer markers through visual observations, absorbance measuring or picture processing through developing the color.When coming visual cancer markers, can wait through visual observations, fluorescence intensity measurement or picture processing and detect cancer markers through fluorescence.Below, be called as miRNA dyeing through colour developing or the cancer markers miRNA that carries out of fluorescence visual.When through the visual cancer markers of for example radioautograph, can wait through for example autoradiographic visual observations or autoradiographic picture processing and detect cancer markers miRNA.

In the present invention, the cancer that assess is not by special restriction.As stated, the example of cancer comprises colorectal carcinoma, the rectum cancer, carcinoma of gallbladder, cancer of the stomach, mammary cancer, white blood disease, carcinoma of the pancreas, liver cancer, brain tumor and osteosarcoma.

In the present invention, sample is a biological sample, and not by special restriction, as long as it is a cell or tissue.The concrete example of sample comprises: tissue and cell around large intestine, rectum, gall-bladder, stomach, breast, hemocyte, lung, brain, bone and the bone; And the cell in blood is such as white corpuscle.

In the present invention, the person under inspection who obtains sample is by special restriction, and can be people for example.Other examples of person under inspection comprise inhuman Mammals, comprising: the primates except the people; Rodents; Dog; And cat.

In the present invention, the cancer markers miRNA in the sample can, for example, directly from sample detection or can detect from RNA indirectly from sample collection.According to the present invention, can for example specify in the zone of expressing cancer markers in the sample.Therefore, preferably, directly detect cancer markers from sample.

To describe at first, below and use the example of coming the method for the cancer markers miRNA the test sample from the RNA of sample collection

Be used for method from sample collection RNA not by special restriction, and, known method can be used.Specifically, can use guanidine-CsCl ultracentrifugation method or AGPC (sour guanidine-phenol-chloroform) etc.Also possibly use commercially available reagent or test kit.

When from product collection RNA, also possibly come to detect indirectly cancer markers miRNA: for example, use collected RNA to synthesize cDNA, detect the cDNA of cancer markers miRNA then from it as template through following manner.

When use utilizes RNA as template synthetic cDNA, can use nucleic acid amplification method for example to carry out the detection of cancer markers miRNA.Nucleic acid amplification method is not by special restriction, and its example comprises polymerase chain reaction (PCR) method, reverse transcription PCR (RT-PCR) method, real-time PCR method and real-time RT-PCR method.Wherein, real-time RT-PCR method preferably.

When for example utilizing nucleic acid amplification method, at first,, and use random primer from the total RNA of sample extraction, with total RNA as the synthetic cDNA of template.Next, use the cDNA that so obtains as template, can the increase primer of cDNA of target cancer markers miRNA of use causes amplified reaction, and detects amplified production.Existence through detecting amplified production or do not exist or the amount of amplified production; Might detect the expression level of the cancer markers miRNA in sample; That is the existence of the expression of cancer markers miRNA or do not exist or the amount of the cancer markers miRNA that in sample, expresses.

The random primer that will in the reaction of synthetic cDNA, use is by special restriction, and, can for example use commercially available random primer.And the primer that will in amplified reaction, use is not restriction extremely, and it can be, for example: can with the cDNA of cancer markers miRNA or with the primer of its complementary sequence hybridization; Perhaps, can with the cDNA of the neighboring area of cancer markers miRNA or with the primer of its complementary sequence hybridization.Can come suitably to design primer based on the base sequence of cancer markers miRNA and general technological general knowledge.Specifically, primer can be, for example: by the cDNA of target cancer markers miRNA or the primer that constitutes with its complementary sequence; Perhaps, by the cDNA of the neighboring area of cancer markers miRNA or the primer that constitutes with its complementary sequence.Preferably; The cDNA of the sequence of primer and for example target cancer markers miRNA or with the cDNA of the neighboring area of its complementary sequence or cancer markers miRNA or with about at least 70% complementation of its complementary sequence; Preferably with they at least 80% complementations; More preferably, be more preferably and they 95% complementations preferred especially and their 100% complementations at least at least with they 90% complementations.

The formation unit of primer is by special restriction, and can use known formation unit.Constitute unitary concrete example and comprise Nucleotide, such as deoxyribonucleotide and ribonucleotide.Constitute unitary example and also comprise PNA (PNAG3) and LNA (lock nucleic acid).Base in constituting the unit is not by special restriction.For example, constitute the unit and can comprise natural base (unartificial base),, perhaps can comprise non-natural base (artificial base) such as A, C, G, T and U.The length of primer is by special restriction, and can be common length.

The method that is used to detect amplified production is by special restriction, and can use currently known methods.When detecting amplified production in real time, preferably, for example, make fluorescent reagent be present in the reaction soln of amplified reaction.The example of fluorescent reagent comprises: with double-strandednucleic acid specificity bonded fluorescent substance; And, insert the fluorescent substance in the double-strandednucleic acid.In amplified reaction, when through when forming double-strandednucleic acid with the extension of template cDNA annealed primer, the fluorescent substance that in reaction soln, exists is attached to or inserts in the double-strandednucleic acid.Then,, might verify the existence of amplified production or do not exist, thus, can verify the existence of target cancer markers miRNA indirectly and do not exist through verifying the fluorescence that is combined or insert the fluorescent substance in the double-strandednucleic acid.And, through measuring the fluorescence intensity of fluorescent substance, can quantize amplified production, thus, quantified goal cancer markers miRNA indirectly.The example of fluorescent reagent comprises SYBR (trade mark) Green.For example, can carry out the detection of using fluorescent reagent through currently known methods.Specifically, under the situation of real-time RT-PCR, can for example use commercially available reagent; Such as SYBR (trade mark) Green PCR Master Mix (trade(brand)name; Perkin-Elmer Applied Biosystem) and, commercially available detector is such as ABI Prim 7900 Sequence Detection System (trade(brand)names; Perkin-Elmer Applied Biosystem), carry out detection according to their handbook.

When detecting amplified production in real time, other examples that are used for detecting the method for amplified production comprise: make the reaction soln at amplified reaction have label probe.Label probe can for example be the probe with fluorescent substance and quencher.The concrete example of such probe comprises TaqMan (trade mark) probe and the circle probe that uses with RNase.When label probe exist singly for example, come the fluorescence of quench fluorescence material through quencher.When label probe forms double-strandednucleic acid, eliminate quenching effect, emitting fluorescence thus.Can for example use such label probe according to currently known methods.

Also possibly use probe to detect the cDNA of cancer markers miRNA.This probe can be, for example, can with the cDNA of cancer markers miRNA or with the primer of its complementary sequence hybridization.In the method, between the cDNA of cancer markers and probe, cause hybridization, and, probe detected with the cDNA hybridization of cancer markers.With the existence of the probe of the cDNA of cancer markers hybridization or do not exist or measure corresponding in the existence of the cancer markers miRNA from the RNA of sample collection or do not exist or measure.Therefore, through detection probes, might detect the existence of the cancer markers miRNA in sample indirectly or do not exist or measure.For example, can carry out the detection of probe through known method, this method can be identical with following method.

In addition, for example, when from sample collection RNA, can directly detect cancer markers miRNA from collected RNA.In this case, detection method can be for example, to use the hybridizing method of probe.As probe, for example, can use can with the probe of cancer markers miRNA specific hybrid.In the method, between cancer markers miRNA and probe, cause hybridization, and detect the probe with cancer markers miRNA hybridization.With the existence of the probe of cancer markers hybridization or do not exist or measure corresponding in the existence of the cancer markers miRNA from the RNA of sample collection or do not exist or measure.Therefore, through detection probes, might detect the existence of the cancer markers miRNA in sample indirectly or do not exist or measure.

The method that is used for detection probes is not by special restriction.The concrete example of detection method for example is, the label probe of applying marking property material mark is used as probe, and can carry out detection through detecting said mark property material.

Serving as a mark property material, for example, can the detectable material of use itself.Said material can for example be substance that show color, the fluorescent substance that sends fluorescence or radioactive substance etc.For example, under the situation of substance that show color, can or not exist based on the existence of colour developing and confirm the existence of substance that show color with intensity or do not exist and measure.Substance that show color can be, for example: the material of colour developing itself; Wait the material of the material that discharges colour developing through enzyme reaction; Perhaps, be converted into the material of substance that show color through enzyme reaction or electrotransfer reaction.For example, under the situation of fluorescent substance, can or not exist based on the existence of fluorescence and confirm the existence of fluorescent substance with intensity or do not exist and measure.This fluorescent substance can be, for example: fluorescigenic material itself; Wait the material that discharges fluorescigenic material through enzyme reaction; Perhaps, be converted into the material of the material that fluoresces through enzyme reaction or electrotransfer reaction.For example, under the situation of radioactive substance, can measure radioactivity level or based on the existence of the image that obtains through radioautograph or do not exist and confirm the existence of mark property material with color density or do not exist and measure through using scintillometer.Use utilizes the example of the hybridizing method that the probe of radioactive substance mark carries out to comprise Northern blotting and microarray analysis.

Mark property material can be, for example, and the mark property material that can enough another kind of reagent detects.The example of such mark property material comprises enzyme, such as SEAP (AP) and horseradish peroxidase (HRP).When mark property material is for example during enzyme; Can add through develop the color with the reaction of enzyme or electronics transmission of following enzyme reaction etc. or the substrate of emitting fluorescence as above-mentioned another kind of reagent, and can detect through with the existence of the colour developing of the reaction of enzyme or fluorescence or do not exist, absorbancy or fluorescence intensity.This substrate is not limited especially, and can suitably be provided with according to the quilts such as kind of enzyme.Concrete example is following: under the situation of for example AP, can use the combination of bromine chlorine oxindole phosphoric acid salt (BCIP) or BCIP and NBT etc.; And, under the situation of HRP, for example, can use 3,3 '-diaminobenzidine four hydrochlorides (DAB) etc.

Can use other examples of the mark property material of another kind of reagent detection to comprise vitamin H and avidin.Wherein, vitamin H is preferred.When using biology usually during label probe, preferably, for example, add above-mentioned enzyme, substance that show color, fluorescent substance or radioactive substance etc. as above-mentioned another kind of reagent, wherein each all has and its bonded avidin.Because the vitamin H as the mark property material of probe combines with avidin, so can detect and avidin bonded enzyme etc. through any of aforesaid method.In addition, above-mentioned another kind of reagent can be, for example, the complex body of avidin and vitamin H and any enzyme etc., it is so-called avidin-vitamin H complex body.This method is so-called ABC (avidin-vitamin H complex body) method.When using such complex body, for example, the avidin in specific complex body can combine with the vitamin H that exists in another complex body, so might increase the molecule number with individual molecule probe bonded enzyme etc.Therefore, has the more highly sensitive detection possibility that becomes.Vitamin H can be the plain verivate of biological example, and avidin can be for example such as the avidin derivative of streptavidin.

And, be used to detect example with the method for the probe of cancer markers miRNA hybridization and comprise that the label probe that wherein will use antigenic mark is used as probe and utilizes the method for antigen-antibody reaction.The concrete example of such method is the method for except the probe of antigenic mark, going back the first antibody of applying marking, and the first antibody of this mark can combine and being labeled property material mark by the specificity conjugated antigen.Another example is to use the method for the SA of first antibody and mark, and this first antibody can the specificity conjugated antigen, and this mark SA can specificity combine first antibody and being labeled property material mark.

When the mark first antibody that for example uses in the former, at first, between the probe of cancer markers miRNA and antigenic mark, cause hybridization.Subsequently, the first antibody of mark via antigen be attached to cancer markers miRNA bonded antigenic mark probe on.Then, detect the mark property material of the mark first antibody of bonding probes.By this way, the probe with cancer markers miRNA hybridization can be detected, thus, cancer markers miRNA can be detected indirectly.The antigenic kind that is used for label probe is not by special restriction, and its example comprises digoxin (DIG).First antibody is by special restriction, and can suitably set according to for example antigenic kind.When antigen is for example during DIG, can use anti-DIG antibody etc.The mark property material of the first antibody of mark is not limited especially, and can be same as described above.

Back a kind of method of the SA of applying marking is so-called sandwich method.In the method, for example, at first, between cancer markers miRNA and antigenic mark probe, cause hybridization.Subsequently, first antibody combines with the antigenic mark probe that combines cancer markers miRNA via antigen.And the SA of mark combines with the first antibody of bonding probes.As a result, the SA of mark combines cancer markers miRNA via first antibody.Then, the mark property material of the SA of certification mark.By this way, the probe of hybridizing can be detected, cancer markers miRNA can be detected indirectly thus with cancer markers miRNA.The antigenic kind and the first antibody of label probe are not limited especially, and can be same as described above.The mark property material of SA is not limited especially, and can be same as described above.

Next, explanation is directly detected the illustrated examples of cancer markers miRNA from sample.

According to the present invention, for example, can specify in the zone of expressing cancer markers miRNA in the sample.Therefore, preferably, directly detect cancer markers miRNA from sample.In this case, preferably, make the sample immobilization.The method that is used to detect cancer markers miRNA is preferably for example used the hybridizing method of probe.Specifically, for example, preferably, detect cancer markers through carrying out in-situ hybridization method for the immobilization sample, and, particularly preferably be, use antigen-antibody reaction to use immunohistochemistry (IHC).According to such method, might for example use cancer markers miRNA to come staining cell matter or nuclear inside.As probe, for example, label probe is preferred, and can use probe same as described above.

Can for example carry out in-situ hybridization method according to known method.And, can use commercially available test kit etc. to carry out in-situ hybridization method according to its working instructions.The example of test kit comprises can be from the RiboMap in situ hybridization test kit (trade(brand)name) of Ventana acquisition.

In in-situ hybridization method for example, carry out the preparation of the section slide glass of sample, pre-treatment, the hybridization of label probe and the detection of hybridization signal of section slide glass.To provide the concrete example of in-situ hybridization method below.Yet, should be noted that to the invention is not restricted to this.

At first, the section slide glass of cell or tissue is prepared as sample.Can prepare the section slide glass through following manner: use fixing of immobilization solution sample, the embedding sample then sample is cut into the sheet with expectation thickness, and the cutting blade that will so obtain places on the slide glass.The example of fixed solution comprises: linking agent, such as formaldehyde and Paraformaldehyde 96; PLP (periodate-lysine-paraformaldehyde); Zamboni solution; LUTARALDEHYDE; And, solidify and precipitation agent, such as ethanol and acetone.Can realize immobilization through for example being selected from for example following any preparation method: for the freezing microtome section of fixed sample not, for the freezing microtome section of immobilization sample with for the paraffin embedding of immobilization sample.Immobilized condition is not limited especially, and, for example, preferably at room temperature carried out fixing 1 hour more preferably at least 6 hours at least.

Next, before hybridization, pre-treatment section slide glass.This pre-treatment can be that for example, deparaffinization, rehydration, retightening, s.t. or use Proteinase K are handled, so that improve the perviousness of label probe.And, for prevent because of the non-specific binding of adding the probe that glycocoll or acetic acid causes so that make Proteinase K lose activity, in also can carrying out with the processing of positive charge etc.

Then, add label probe,, and carry out the detection of hybridization signal with the mode of the mark property material that is suitable for label probe so that cause hybridization to pretreated section slide glass.

The label probe amount of adding is by special restriction, and can suitably be provided with respect to the ratio of the probe that will be used as a whole etc. according to the kind of for example mark property material, label probe.As a specific example, the label probe amount that add can but for example never be limited in position in the hybridizing method normally used each single slide glass 1 to 1000ng.

The condition of hybridization is not by special restriction.Concrete example is following.Thermally denature before hybridization is handled preferably and was carried out 1 to 60 minute at 50 ℃ to 100 ℃, more preferably carries out 5 to 10 minutes at 60 ℃ to 95 ℃.And hybridization was preferably carried out 1 to 36 hour at 40 ℃ to 80 ℃, more preferably carried out 4 to 24 hours at 45 ℃ to 70 ℃.

Next, carry out the detection of hybridization signal.The method that is used for detection signal is by special restriction, and can be for example suitably confirms according to the kind of the mark property material of the SA of the first antibody of aforesaid label probe and mark or mark.When for example detecting colour developing or fluorescence, can or not exist or detect the existence of the cancer markers miRNA in sample or do not exist or measure based on the existence of colour developing or fluorescence based on the density of the color that is shown or intensity of fluorescence.Can for example check colour developing or fluorescence through visual observations or picture processing.When mark property material is radioactive substance, can for example uses radioautograph, and can or not exist or color density detects the existence of the cancer markers miRNA in sample or do not exist or measure based on autoradiographic existence.Can check autoradiographic existence or do not exist or color density through for example visual observations or picture processing.Picture processing is by special restriction, and can use known system or known software to be performed.

When in the detection of cancer markers miRNA, using probe as stated, the sequence of probe is not by special restriction.The example of probe comprises the probe that can specificity combines any above-mentioned cancer markers miRNA.For example, commercially available product can be used as probe, or can prepare probe voluntarily.Can come the suitably sequence of designing probe based on base sequence and the general technological general knowledge of for example above-mentioned cancer markers miRNA.Specifically, for example, probe can be by with detect probe that target cancer markers miRNA complementary sequence constitutes or the probe that comprises said complementary sequence.Preferably, for example, the sequence of probe and about at least 70% complementation of target cancer markers miRNA, more preferably with its at least 90% complementation, and preferred especially and its 100% complementation.

The formation unit of probe is by special restriction, and for example, can use known formation unit.Constitute unitary concrete example and comprise Nucleotide such as deoxyribonucleotide and ribonucleotide; PNA; And, LNA.The example of LNA comprises BNA (nucleic acid of bridge joint), such as 2 ', and 4 '-bridge joint nucleic acid.Base in constituting the unit is not by special restriction.They can be, for example, and such as the natural base of VITAMIN B4, guanine, cytosine(Cyt), thymus pyrimidine and uridylic, or can the natural base of right and wrong.The length of probe is by special restriction, and is for example 10 to 100 bases, preferably 15 to 40 bases.

The concrete example of probe is shown below.Yet, should be noted that to the invention is not restricted to this.Other examples of probe comprise by the polynucleotide that constitute with the base sequence complementary sequence shown in the SEQ ID NO:5.

Hsa-miR-92a detection probes (SEQ ID NO:5)

5’-acaggccgggacaagtgcaata-3’

Appraisal procedure of the present invention is evaluated at the possibility of the cancer in the sample based on the expression level of the cancer markers miRNA that detects in the above described manner in appraisal procedure.

The expression level of cancer markers can be, for example, and the existence that cancer markers miRNA expresses in sample or do not exist or the amount of the cancer markers miRNA that in sample, expresses.Expression amount can be, for example, and the actual amount of miRNA or the value relevant with the actual amount of miRNA.The latter's example comprises the signal value that when detecting cancer markers miRNA, obtains.For example, can suitably confirm signal value according to being used to detect the method for miRNA and the kind of the detector that is used for detected signal value etc.When detection method is when for example comprising any nucleic acid amplification method of PCR method such as real-time RT-PCR method, can signal value be expressed as number of copies (copy/μ l) of for example per 1 μ l etc.And, be described below, when using when having carried out miRNA colored graph picture (below be called as " miRNA colored graph picture "), for example, the value (brightness) of colour developing or fluorescence or saturation ratio are corresponding to signal value.

In appraisal procedure of the present invention; When the cancer markers miRNA that detects as stated in the immobilization sample; Preferably; For example, provide the visual section slide glass of cancer markers miRNA (below be also referred to as " the visual section slide glass of cancer markers miRNA ") and use the painted section slide glass of HE (phenodin & Yihong), and they are contrasted each other.

Current, for the canceration of confirming cell or tissue by the pathologist, use HE colored graph picture (below be called " HE colored graph picture ").Yet HE dyeing has for example following problems: the assessment for the border pathological change becomes difficult; And when the quantity of the cell of checking was 1 or 2, it is difficult that assessment becomes.For this reason, carry out the HE dyeing of cancer markers miRNA and visual with respect to the section slide glass that draws from same sample, and the slide glass of will cutting into slices contrasts each other.Then, through will based on HE dye specified tumor region with for the contrast of visual cancer markers miRNA male zone, might confirm the possibility of cancer with higher safety.For example there is cancer markers in term " positive " expression, and for example there is not cancer markers in term " feminine gender " expression, and the amount of cancer markers is lower than detectability.

Through with the visual section slide glass of cancer markers miRNA with carry out the painted section slide glass of HE contrast confirm the cancer possibility method not by special restriction, and can for example carry out in the following manner.The mode stated of chatting face to face at first, is later on carried out the HE dyeing of miRNA and visual with respect to adjacent section slide glass.About the painted section slide glass of HE, for example, observe to confirm tumor region through using microscope etc.Then, with painted section slide glass of HE and the visual section slide glass contrast of cancer markers.As a result of, for example, when the tumor region in the painted section slide glass of HE is consistent with the cancer markers positive region in the visual section slide glass of cancer markers, can confirm that this tumor region and cancer markers positive region are carcinous.

Not only carrying out the visual of cancer markers miRNA for sample but also carrying out under the painted situation of HE; Preferably; Being used for section slide glass that cancer markers miRNA detects is the adjacent section that the sample from immobilization and embedding cuts out with being used for the painted section slide glass of HE; Specifically, preferably, adjacent section is a serial section.This allows map picture more accurately.

Preferably, look like to carry out the contrast between visual cancer markers miRNA and HE dyeing through for example map.That is, preferably, the visual image of cancer markers (below be also referred to as " cancer markers visual image ") is provided for the immobilization sample and uses HE colored graph picture, and they are contrasted each other.For example, through following manner image is provided: through CCD (ccd image sensor) or scanning device etc. will be digital picture through the visual immobilization sample of cancer markers miRNA with through the HE visual immobilization sample changeover that dyes.

Through using aforesaid image, can be more easily and accurately carry out confirming of contrast and cancer possibility between visual cancer markers miRNA and HE dyeing.And, through the accumulative image data, further assess the possibility that becomes more reliably.

Through cancer markers miRNA visual image and HE colored graph are looked like to contrast confirm the cancer possibility method not by special restriction, and can for example carry out in the following manner.At first, carry out the visual of HE dyeing and cancer markers miRNA, therefore HE colored graph picture and cancer markers visual image are provided for aforesaid contiguous slices slide glass.Then, specify in the tumor region in the HE colored graph picture, and, HE colored graph picture and cancer markers visual image are contrasted each other.As a result, when tumor region in HE colored graph picture and the cancer markers miRNA positive region in the cancer markers visual image are consistent each other, can confirm that this tumor region and cancer markers positive region are carcinous.And, when in HE colored graph picture, not finding tumor region and cancer markers visual image to be feminine gender, can confirm that sample is not carcinous for cancer markers miRNA.And; When tumor region in HE colored graph picture and the cancer markers positive region in the cancer markers visual image are inconsistent; Leave last judgement for pathologist, and these data can be stored as cumulative data, so that for example prepare to be used for judgement from now on.

The present invention discloses, and is for example increasing with in the cell of canceration and the tissue as the expression level of the has-miR-92a of cancer markers miRNA and has-miR-494.Can explain that from this situation for example, the expression level of each cancer markers miRNA significantly increases: make comparisons before with outbreak in cancer in cancer outbreak back; Make comparisons at preclinical phase and before preclinical phase; Make comparisons in the clinical phase and before the clinical phase; Make comparisons before in the early stage with in the early stage; And, back and just interim making comparisons in the early stage.Therefore, appraisal procedure of the present invention can for example comprise in appraisal procedure: for example, come to confirm based on the expression level of cancer markers miRNA the possibility of cancer through at least a method that is selected from following (1), (2) and (3).In the present invention, term " normal person " expression for example, is not confirmed as and has suffered from the cancer that will assess or had the evaluation object that he has suffered from the possibility of cancer.On the other hand, term " patient " expression for example, is not confirmed as the evaluation object of having suffered from the cancer that will assess.

(1) will make comparisons at the expression level of the cancer markers miRNA in person under inspection's the sample and the expression level of the cancer markers miRNA in normal person's sample; And when the expression level in the person under inspection is higher than the expression level in normal person, confirm that this person under inspection has the high likelihood of cancer.

(2) will make comparisons at the expression level of the cancer markers miRNA in person under inspection's the sample and the expression level of the cancer markers miRNA in normal person's sample; And when the expression level in the person under inspection uprises than the expression level in normal person relatively, confirm that the cancer in the person under inspection is made progress relatively.

(3) will make comparisons at the expression level of the cancer markers miRNA in the expression level of the cancer markers miRNA in person under inspection's the sample and each cancer patients's of different progressive stages the sample; And, confirm that the cancer in the person under inspection is in identical progressive stage with cancer in the patient who shows identical or similar expression level.

For example, can use the expression level of confirming normal person's cancer markers miRNA method (1) and (2) from the sample of normal person's collection.And, can confirm the expression level of cancer patients's cancer markers miRNA in method (3) through for example following manner:, and use the sample of collecting from the patient in corresponding progressive stage according to the progressive stage patient that classifies.In (3), for example can confirm normal person and patient's expression level in advance in method (1), and confirm they needn't assess the time at every turn.In (3), for example, the kind of normal person and patient's the sample preferably kind with person under inspection's sample is identical in method (1).And for example, normal person and patient's sample preferably prepares with person under inspection's sample in an identical manner and under identical condition.The expression level of normal person or each patient's cancer markers miRNA can be, for example, from the value of single normal person or single patient acquisition, perhaps can be through the value of statistical method from a plurality of normal persons or a plurality of patients' expression level calculating.

In aforesaid method (1), when the expression level in the person under inspection is higher than the expression level in normal person, can confirm that the person under inspection has the high likelihood of cancer.On the other hand, when the expression level in the person under inspection is equal to or less than the expression level in normal person, can confirm that the person under inspection has the low possibility of cancer.The possibility that the possibility of cancer also can be called as possibility that canceration has for example taken place or the person under inspection possibly suffer from cancer.

In aforesaid method (2), when the expression level in the person under inspection becomes higher relatively than the expression level in normal person, can confirm that the cancer in the person under inspection is made progress relatively.And even when the expression level in the person under inspection is higher than the expression level in normal person, but the difference between expression level is when diminishing, and can confirm not progress relatively of cancer.

In method (3), for example, confirm expression level each cancer patients of different progressive stages.Through the comparison between each patient's expression level and the expression level in the person under inspection, can not only assess person under inspection's canceration possibility, and the progressive stage of assessment of cancer.

In (3), when expression level that will be in the person under inspection is made comparisons with normal person or each patient's expression level, for example, can confirm significant difference therebetween in method (1) through statistical method such as t check, F check or chi square test.

According to such appraisal procedure, for example,, might judge the person under inspection of high likelihood with high reliability, and such judgement is being difficult traditionally with cancer for the person under inspection who suffers from the preclinical phase cancer.And, for example, also can high reliability judge the cancer progression phase.Therefore, in the prevention or treatment of cancer, can high reliability obtain for example important information in the strategy of decision medication, operation etc.

And for example, appraisal procedure of the present invention can be come assessment of cancer through the dose rate that contains that calculates the stained positive cell.Term " stained positive cell " expression for example has the cell that cancer markers and the dyeing through cancer markers are colored.In this case, in this appraisal procedure of the present invention, for example, detect in cancer markers and to obtain in the step to have carried out cancer markers colored graph picture (below be called " cancer markers colored graph picture ") about the immobilization sample.Appraisal procedure of the present invention preferably further comprises:

HE colored graph picture obtains step, obtains the HE colored graph picture about said immobilization sample;

The information acquisition step, the information of the tumor region of acquisition in said HE colored graph picture;

The coupling step is calculated the matched position that obtains said HE colored graph picture that obtains in the step and the said cancer markers colored graph picture that in said cancer markers detection step, obtains at the HE image;

Given step; The information of the said matched position that is based on the information of the said tumor region in the said HE colored graph picture that obtains in the said information acquisition step and in said coupling step, calculates specifies in the tumor region in the said cancer markers colored graph picture; And

Stained positive cell detection step is based on the information of the said tumor region in the specified said cancer markers colored graph picture in the said given step, detects the stained positive cell in the said tumor region in said cancer markers colored graph picture.

Said stained positive cell detection step can be a calculation procedure for example, and the information that is based on the said tumor region in the specified said cancer markers colored graph picture in the said given step is calculated the stained positive cell containing ratio at tumor region described in the said cancer markers colored graph picture.Said cancer markers colored graph picture can be for example, to have the image through colour developing or luminous visual cancer markers miRNA.

As stated, for example, can carry out the cancer assessment that comprises with the contrast of HE colored graph picture through following cancer pathology diagnostic imaging support method.Can realize this method through for example operating cancer pathology image diagnosis support system, cancer pathology diagnostic imaging support program or cancer pathology diagnostic imaging support equipment---wherein each will be described below---.To describe in detail this below.

< assessment reagent >

As stated; Assessment reagent of the present invention is the assessment reagent that will in appraisal procedure of the present invention, use; And it is characterized in that it comprises the reagent that is used to detect cancer markers of the present invention; That is, be used to detect the miRNA detection reagent of at least one miRNA that is selected from has-miR-92 and has-miR-494.According to such assessment reagent, might carry out appraisal procedure of the present invention easily.

As stated, the invention is characterized in that with has-miR-92 and has-miR-494 at least one detects as cancer markers miRNA, and never restriction is used to detect the method for these miRNA.The miRNA detection reagent that only need in assessment reagent of the present invention, comprise can detect any of these miRNA, and never limit reagent for example kind, become to grade.And those skilled in the art can be provided for the detection reagent of these cancer markers miRNA based on general technological general knowledge.

The miRNA detection reagent is by special restriction, and its example comprise can be with cancer markers miRNA, such as the probe of any one hybridization of aforesaid cancer markers miRNA.Probe can be aforesaid label probe.And, kind of the mark property material that uses according to the method that is used for detecting miRNA with at label probe etc., the miRNA detection reagent can further comprise other reagent.

Assessment reagent of the present invention can further comprise enzyme, damping fluid, washings, lysate, dispersion liquid and diluent etc. according to for example being used to detect the method for miRNA.And the form of assessment reagent of the present invention is not by special restriction.For example, it can be with the wet type reagent of liquid state or with the dry type reagent of dryness.

< assessment test kit >

Assessment test kit of the present invention is the assessment test kit that will in appraisal procedure of the present invention, use as stated, and it is characterized in that it comprises the miRNA detection reagent that is used to detect at least one miRNA that is selected from has-miR-92 and has-miR-494.The example of miRNA detection reagent comprises aforesaid assessment reagent of the present invention.According to such assessment test kit, can carry out appraisal procedure of the present invention easily.

The form of assessment test kit of the present invention is not by special restriction.It can be the wet type test kit of for example liquid state or the dry type test kit of dryness.For example, the corresponding reagent in assessment test kit of the present invention can provide respectively, and when using this test kit, uses together, or can before using this test kit, mix.Assessment test kit of the present invention can for example comprise working instructions.

The invention provides system, program, method and apparatus, every kind all is used to support the cancer diagnosis based on pathological image, and its example comprises the first and second following embodiments.

First embodiment

The invention provides a kind of cancer pathology image diagnosis support system (below be called " image diagnosis support system ") that is used to support based on the cancer diagnosis of pathological image, comprising:

Image obtains the unit, and it obtains HE colored graph picture and cancer markers colored graph and looks like to be used as the pathological image that will diagnose;

The information acquisition unit, it obtains the information of the tumor region in said HE colored graph picture;

Matching unit, it calculates through said image and obtains the said HE colored graph picture of unit acquisition and the matched position of said cancer markers colored graph picture;

Designating unit, the information of the said matched position that it calculates based on the information of tumor region described in the said HE colored graph picture that is obtained by said information acquisition unit with by said matching unit specifies in the tumor region in the said cancer markers colored graph picture; And

Stained positive cell detection unit, it detects the stained positive cell in tumor region described in the said cancer markers colored graph picture based on the information by the said tumor region in the specified said cancer markers colored graph picture of said designating unit.

Said stained positive cell detection unit can be; For example, based on by the information of tumor region described in the specified said cancer markers colored graph picture of said designating unit, calculate the computing unit (below identical) of the containing ratio of the stained positive cell in the said tumor region in said cancer markers colored graph picture.

Preferably, said computing unit further calculates staining power except said stained positive cell containing ratio.Said staining power can be, for example, and at the staining power of tumor region described in the said cancer markers colored graph picture.

Preferably, said image diagnosis support system of the present invention further comprises:

The unit is accepted in input, the input of information that its acceptance is used to specify the pathological image that will diagnose and the information that is used to specify inspect-type; And

The dyeing image data base, it stores said HE colored graph picture and said cancer markers colored graph picture, wherein

Said image obtains the unit and obtains said HE colored graph picture and said cancer markers colored graph picture based on said appointed information from said dyeing image data base.

In said image diagnosis support system of the present invention; Preferably; The said information that is used to specify the said pathological image that will diagnose is the image identifier of said HE colored graph picture, and said image obtain the unit from said dyeing image data base obtain to have said image identifier said HE colored graph picture and with said HE colored graph as adjacent said cancer markers colored graph picture.

In said image diagnosis support system of the present invention; Preferably; The said information that is used to specify the said pathological image that will diagnose is the image identifier of said cancer markers colored graph picture, and said image obtain the unit from said dyeing image data base obtain to have said image identifier said cancer markers colored graph picture and with said cancer markers colored graph as adjacent said HE colored graph picture.

In said image diagnosis support system of the present invention; Preferably; The said information that is used to specify the said pathological image that will diagnose is person under inspection's identifier of diagnosis target; And said image obtains the unit and obtains said HE colored graph picture and said cancer markers colored graph picture from said dyeing image data base, and wherein each has object identifier.

In said image diagnosis support system of the present invention; Preferably; Said dyeing image data base is also stored the said information of tumor region described in the said HE colored graph picture, and said information acquisition unit is from the said information of the said tumor region of said dyeing image data base acquisition said HE colored graph picture.

Preferably; Said image diagnosis support system of the present invention further comprises: the tumor region computing unit; It calculates the said tumor region in the said HE colored graph picture that is obtained by said image acquisition unit; Wherein, said information acquisition unit acquisition is by the information of the said tumor region in said HE colored graph picture of said tumor region computing unit calculating.

The unit is accepted in input, and the slide identifier of the slide glass that its acceptance will be diagnosed accords with and be used to specify the input of the information of inspect-type;

The slide glass DB, it stores said slide glass; And

Slide glass obtains the unit, and it obtains to have the said slide glass that said slide identifier accords with from said slide glass DB, wherein

Said image obtains the unit through said slide glass being obtained the said slide glass imaging that the unit obtains, and obtains said HE colored graph picture and said cancer markers colored graph picture.

The invention provides a kind of image diagnosis support system that is used to support to carry out cancer diagnosis, comprising: terminal and server based on pathological image.Said terminal and said server can be via at the network of communication of said system outer setting and be connected to each other.Said terminal comprises: the end side transmitting element, and it will send to said server via said network of communication in the information in the said terminal; And, the end side receiving element, it receives the information of sending from said server via said network of communication.Said server comprises: the server side transmitting element, and it will send to said terminal via said network of communication in the information in the said server; The server side receiving element, it receives the information of sending via said network of communication from said terminal; Image obtains the unit, and its acquisition HE colored graph picture and cancer markers colored graph picture are as the pathological image that will diagnose; The information acquisition unit, it obtains the information of the tumor region in said HE colored graph picture; Matching unit, it calculates through said image and obtains the said HE colored graph picture of unit acquisition and the matched position of said cancer markers colored graph picture; Designating unit, the information of the said matched position that it calculates based on the said information of the said tumor region in the said HE colored graph picture that is obtained by said information acquisition unit with by said matching unit specifies in the tumor region in the said cancer markers colored graph picture; And; Stained positive cell detection unit, its information that is based on by the said tumor region in the specified said cancer markers colored graph picture of said given step detects the stained positive cell in the said tumor region in said cancer markers colored graph picture.In this image diagnosis support system; The information of sending said pathological image to said server side receiving element from said end side transmitting element; And, the information of sending the said stained positive cell that detects the said stained positive cell detection unit by said server from said server side transmitting element to said end side receiving element.

In said image diagnosis support system of the present invention; Said stained positive cell detection unit can be; For example, based on containing ratio by the stained positive cell in the information of the said tumor region in the specified said cancer markers colored graph picture of said designating unit, the said tumor region in said cancer markers colored graph picture.In this case, for example, send the information of the said stained positive cell containing ratio that the said computing unit by said server calculates to said end side receiving element from said server side transmitting element.

The present invention also provides a kind of server of using at said image diagnosis support system of the present invention of being used for.Said server comprises: the server side transmitting element, and it sends to said terminal with the information in the said server via said network of communication; The server side receiving element, it receives the information of sending via said network of communication from said terminal; Image obtains the unit, and its acquisition HE colored graph picture and cancer markers colored graph picture are as the pathological image that will diagnose; The information acquisition unit, it obtains the information of the tumor region in said HE colored graph picture; Matching unit, it calculates through said image and obtains the said HE colored graph picture of unit acquisition and the matched position of said cancer markers colored graph picture; Designating unit, the information of the said matched position that it calculates based on the said information of the said tumor region in the said HE colored graph picture that is obtained by said information acquisition unit with by said matching unit specifies in the tumor region in the said cancer markers colored graph picture; And; Stained positive cell detection unit, its information that is based on the said tumor region in the specified said cancer markers colored graph picture of said given step detects the stained positive cell in the said tumor region in said cancer markers colored graph picture.

In server of the present invention; Said stained positive cell detection unit can for example be a computing unit; It calculates the containing ratio of the stained positive cell in the said tumor region in said cancer markers colored graph picture based on the information by the said tumor region in the specified said cancer markers colored graph picture of said designating unit.

The present invention also provides a kind of terminal of in said image diagnosis support system of the present invention, using.Said terminal comprises: the end side transmitting element, and it sends to server with the information in the said terminal via network of communication; And, the end side receiving element, it receives the information of sending via said network of communication from said server.In said terminal; Send the information of pathological image to said server side receiving element from said end side transmitting element; And, the information of sending the stained positive cell that detects by said stained positive cell detection unit from said server side transmitting element to said end side receiving element.

And, for example, can be from said server side transmitting element to the information of said end side receiving element transmission by the said stained positive cell containing ratio of the said computing unit calculating of said server.

The present invention also provides a kind of cancer pathology diagnostic imaging support method (below be called " diagnostic imaging support method ") that is used to support carry out based on pathological image cancer diagnosis, may further comprise the steps:

Obtain HE colored graph picture and cancer markers colored graph picture as the pathological image that will diagnose;

The information of the tumor region of acquisition in said HE colored graph picture;

Matching image is to calculate the said HE colored graph picture that obtains to obtain in the step at said image and the matched position of said cancer markers colored graph picture;

The information of the said matched position that is based on the said information of the said tumor region in the said HE colored graph picture that is obtained by said information acquisition step and is calculated by said coupling step specifies in the tumor region in the said cancer markers colored graph picture; And

The information that is based on by the said tumor region in the specified said cancer markers colored graph picture of said given step detects the stained positive cell in the said tumor region in said cancer markers colored graph picture.

In diagnostic imaging support method of the present invention; Said stained positive cell detection step can for example be a calculation procedure, and the information that is based on the said tumor region in the specified said cancer markers colored graph picture of said given step is calculated the containing ratio of the stained positive cell in the said tumor region in said cancer markers colored graph picture (below identical).

In diagnostic imaging support method of the present invention, preferably, in said calculation procedure, except said stained positive cell containing ratio, further calculate staining power.Said staining power can for example be the staining power in the said tumor region in said cancer markers colored graph picture.

In diagnostic imaging support method of the present invention, preferably, obtain for example, to specify the information of the said pathological image that will diagnose to obtain said HE colored graph picture and said cancer markers colored graph picture in the step based on being used at said image.Preferably, obtain said HE colored graph picture and said cancer markers colored graph picture from the said dyeing image data base of for example storing said HE colored graph picture and said cancer markers colored graph picture.

In diagnostic imaging support method of the present invention, preferably, the said information that is used to specify the said pathological image that will diagnose is the image identifier of said HE colored graph picture.Obtain in the step at said image, preferably, for example, from said dyeing image data base obtain to have said image identifier said HE colored graph picture and with said HE colored graph as adjacent said cancer markers colored graph picture.

In said diagnostic imaging support method of the present invention, preferably, the said information that is used to specify the said pathological image that will diagnose is the image identifier of said cancer markers colored graph picture.Obtain in the step at said image, preferably, for example, from said dyeing image data base obtain to have said image identifier said cancer markers colored graph picture and with said cancer markers colored graph as adjacent said HE colored graph picture.

In diagnostic imaging support method of the present invention, preferably, the said information that is used to specify the said pathological image that will diagnose is person under inspection's identifier of diagnosis target.Obtain in the step at said image, preferably, for example, obtain said HE colored graph picture and said cancer markers colored graph picture with person under inspection's identifier from said dyeing image data base.

In diagnostic imaging support method of the present invention, preferably, said dyeing image data base is also stored the said information of the said tumor region in the said HE colored graph picture.In said information acquisition step, preferably, for example, from the said information of the said tumor region of dyeing image data base acquisition said HE colored graph picture.

Preferably, said diagnostic imaging support method of the present invention further comprises the step of the said tumor region of calculating in the said HE colored graph picture that said image obtains to obtain in the step.In said information acquisition step, preferably, the information of the said tumor region in the said HE colored graph picture that obtains in said tumor region calculation procedure, to calculate.

The present invention also provides a kind of cancer pathology diagnostic imaging support program (below be called " diagnostic imaging support program "), is used to support the cancer diagnosis based on pathological image.Said diagnostic imaging support program is characterised in that it can be so that computingmachine be carried out image diagnosing method of the present invention.

Said diagnostic imaging support program of the present invention makes computingmachine carry out, for example:

Image obtains step, obtains HE colored graph picture and cancer markers colored graph picture as the pathological image that will diagnose;

The coupling step is calculated the said HE colored graph picture that said image obtains to obtain in the step and the matched position of said cancer markers colored graph picture;

Given step is based on the said information of the said tumor region in said HE colored graph picture that obtains in the said information acquisition step and the information of the said matched position that in said coupling step, calculates specifies in the tumor region in the said cancer markers colored graph picture; And

Stained positive cell detection step, the information that is based on the said tumor region in the specified said cancer markers colored graph picture of said given step detects the stained positive cell in the said tumor region in said cancer markers colored graph picture.

In said diagnostic imaging support program of the present invention; Said stained positive cell detection step can for example be a calculation procedure, and its information that is based on the said tumor region in the specified said cancer markers colored graph picture of said given step is calculated the containing ratio of the stained positive cell in the said tumor region in said cancer markers colored graph picture.

According to above-mentioned supporting system, support method and support program, also possibly for example in the end obtain stained positive cell containing ratio as quantized value.

Below, will specifically go up reference implementation scheme 1A to 1D and describe first embodiment of the present invention.Below, cancer markers colored graph picture also is called as " miRNA colored graph picture ".Should be noted that the present invention never is limited to these embodiments.

(embodiment 1A)

Figure 19 illustrates the block diagram that has according to the example of the configuration of the diagnostic imaging support equipment of image diagnosis support system of the present invention.As shown in Figure 19, diagnostic imaging support equipment 190 comprises handling part 191 and storage part 192.Diagnostic imaging support equipment 190 is configured to make it to be connected to the CCD 194 that is connected with microscope 193 at handling part 191 places, and is connected to scanning device 195 and indicating meter 196.Diagnostic imaging support equipment 190 comprises, for example, and CPU (central processing unit), RAM (RAS), input part, driving mechanism, input-output interface (I/F) and communication bus etc.CPU control entire image diagnosis supportive device.Through the computer program of each the unitary function for example be used for being provided at CPU is installed, might be configured in each unit in the diagnostic imaging support equipment 190, thereby realize diagnostic imaging support equipment 190.And, also can realize the operation of diagnostic imaging support equipment 190 through being equipped with circuit block, this circuit block comprises hardware component, such as LSI (integrated on a large scale), the computer program that is used to realize each unitary function has been installed wherein.Such computer program can be the form of the recording medium etc. of storage computer program.The example of recording medium comprise HDD, FD, CD-ROM (CD-R, CD-RW), MO, DVD and storage card.The example of storage part 192 comprises ROM, HDD and HD.HDD for example controls under the control of CPU and reads and write data about HD.The HD data that for example storage is write under the control of HDD.

Indicating meter 196 for example shows the various information such as image, data and file.The example of input part comprises keyboard and mouse.For example, scanning device 195 scans above-mentioned section slide glass, and converts image optics into electrical signal.For example, will the cut into slices micro-image of slide glass of CCD 194 converts electrical signal into.

Diagnostic imaging support equipment 190 can for example can visit diagnostic imaging support equipment 190 outer setting, be used to accumulate dyeing image data base about the information of colored graph picture.In this case, diagnostic imaging support equipment 190 can for example be connected to the dyeing image data base via order wire.

Provide the illustrated examples of image diagnosis support system of the present invention below.Figure 20 is the block diagram that schematically shows according to the configuration of the image diagnosis support system of this embodiment.Should be noted that the present invention never is limited to this embodiment.

As shown in Figure 20, the image diagnosis support system of this embodiment comprises: image obtains unit 2001, and it obtains HE colored graph picture and miRNA colored graph picture; Information acquisition unit 2002, it obtains the information of the tumor region in HE colored graph picture; Matching unit 2003, it calculates by image and obtains the HE colored graph picture of unit acquisition and the matched position of miRNA colored graph picture; Designating unit 2004, the information of the matched position that it calculates based on the information of the tumor region in the HE colored graph picture that is obtained by the information acquisition unit with by matching unit specifies in the tumor region in the miRNA colored graph picture; And, computing unit 2005, it is based on the containing ratio that is calculated the stained positive cell in the tumor region of miRNA colored graph picture by the information of the tumor region in the specified miRNA colored graph of the designating unit picture.

Image diagnosis support system may further include dye level and confirms the unit, and it confirms the miRNA dye level of miRNA colored graph picture.Dye level confirms that the unit for example confirms the dye level in the appointment tumor region of miRNA colored graph picture.

The example of such system is included in the diagnostic imaging support equipment shown in Figure 19.Each constitutes the unit and can for example be made up of the functional block that realizes through the CPU execution preset program by computingmachine.Therefore, for example, each constitutes the unit and not necessarily is provided as hardware component, and they may be provided in network system.Unless otherwise indicated, the image diagnosis support system of this embodiment is identical with the for example image diagnosis support system according to following embodiment 1B to 1E.

With reference to Figure 21, with an example of the treatment scheme in the image diagnosis support system that is described in this embodiment.Figure 21 is the schema that this treatment scheme is shown.This processing is the example according to image diagnosing method of the present invention, and can be for example by execution such as image diagnosis support system of the present invention or diagnostic imaging support programs.

At first, obtain HE colored graph picture and miRNA colored graph picture (step S2101).For example, image can be used as the electrical signal that is produced by the conversion such as the image-pickup device of scanning device or CCD and obtains.

Next, about HE colored graph picture, obtain the information (S2102) of tumor region.The information of the tumor region in HE colored graph picture can be for example to wait the information of judging by the doctor, maybe can be the information of being calculated by currently known methods.

Next, through one of HE colored graph picture and miRNA colored graph picture is overlapping and calculate matched position and carry out coupling with another, (S2103).Then, based on the information of the tumor region in the HE colored graph picture and the information of the matched position that so obtains, calculate the tumor region in miRNA colored graph picture.That is, calculate with HE colored graph picture in the zone of the corresponding miRNA colored graph picture of tumor region, then, tumor region (S2104) is appointed as in this zone.Below, based on the information of HE colored graph picture and the tumor region in the specified miRNA colored graph picture also is called as " based on the tumor region of HE colored graph picture ".

Subsequently, confirm the miRNA dye level (S2105) of the tumor region in the miRNA colored graph picture of confirming based on HE colored graph picture.About the miRNA dye level, for example, preferably, the carries out image stdn is because painted degree is according to the kind of dyeing processing, temperature, probe kind or substance that show color or fluorescent substance etc. and different between slide glass.At this moment, in miRNA colored graph picture, can come to confirm miRNA dye level in a like fashion except based on the zone the tumor region of HE colored graph picture.

Then, in such as aforesaid dyeing image data base, be accumulated in the dyeing graphicinformation (S2106) that obtains in this step.

Next, the miRNA dye level based on so confirming detects the tumor region (S2107) in miRNA colored graph picture again.That is, about the dye level of miRNA colored graph picture, confirm it whether reach the expression tumour cell level or less than the expression tumour cell level.Then, the tumor region based on the miRNA dye level is appointed as in the zone of a kind of dye level before showing.As a result, when based on the tumor region of HE colored graph picture with based on the tumor region of miRNA dye level when consistent, be confirmed as based on the tumor region of miRNA dye level and detect the target area.And, when this zone is not confirmed as tumor region and this zone in the HE colored graph picture and is confirmed as the tumor region based on the miRNA dye level, be confirmed as based on the tumor region of miRNA dye level and detect the target area.

For example, can confirm the dye level of expression tumour and less than the threshold value between the dye level of representing tumour through the miRN staining power that detects a plurality of levels in the miRNA colored graph picture.Therefore, for example,, can confirm this and not represent this dyeing indication tumour cell even when non-tumor cell only slightly dyes.For example, the data about threshold value also are stored in the dyeing image data base as the information about the colored graph picture.

The detection target area of confirming in the above described manner is outputted as definite cancerous area.Perhaps,, calculate stained positive cell containing ratio in detecting the target area, and output calculation result.

When in the dyeing image data base, accumulating the information of miRNA colored graph picture, for example, possibly not carry out confirming the tumor region in miRNA colored graph picture based on this DB with the coupling of HE colored graph picture yet.

(embodiment 1B)

Fig. 1 is the block diagram that illustrates according to the example of image diagnosis support system of the present invention.This system is used to support the system based on the cancer diagnosis of pathological image.This system comprises: image obtains the unit, and its acquisition HE colored graph picture and miRNA colored graph picture are as the pathological image that will diagnose; The information acquisition unit, it obtains the information of the tumor region in HE colored graph picture; Matching unit, it calculates by image and obtains the HE colored graph picture of unit acquisition and the matched position of miRNA colored graph picture; Designating unit, the information of the matched position that it calculates based on the information of the tumor region in the HE colored graph picture that is obtained by the information acquisition unit with by matching unit specifies in the tumor region in the miRNA colored graph picture; And, computing unit, it is based on the containing ratio that is calculated the stained positive cell in the tumor region of miRNA colored graph picture by the information of the tumor region in the specified miRNA colored graph of the designating unit picture.

In more detail, the system of this embodiment comprises input unit 111, take-off equipment 112, dyeing image data base 113, treatment unit 120 and storing device 130.

Dyeing image data base 113 storages: one or more HE colored graph pictures; The miRNA colored graph picture of the sample of the serial section adjacent with the sample (section slide glass) of said HE colored graph picture; Sample neighbor information about above-mentioned HE colored graph picture and above-mentioned miRNA colored graph picture; And, the tumor region information of confirming from above-mentioned HE dyeing image calculation or by the doctor etc.

Each image has person under inspection's identifier, and is relevant with each image through it about person under inspection's relevant information.For example, as shown in Figure 2, dyeing image data base 113 comprises: person under inspection's identifier 201 is used for identifying uniquely the person under inspection; Image identifier 202; Dyeing information 203; View data 204; Sample neighbor information 205; And the HE colored graph is as tumor region information 206.

Image identifier 202 is the identifiers that are used to discern a plurality of pathological images of each person under inspection.Dyeing information 203, view data 204 and tumor region information 206 are distinguished through those of image identifier 202 and other images mutually.The dyeing information of each dyeing information 203 indicating image, and the example of dyeing information comprises about the painted information of HE with about the painted information of cancer markers miRNA.Each view data 204 storing image data.Sample neighbor information 205 is stored the correspondence relation through using image identifier 202.The HE colored graph is stored from HE dyeing image calculation or by definite tumor region information such as doctors as tumor region information 206.Simultaneously, can so that the HE colored graph as tumor region information 206 corresponding to image identifier 202, and separately storage.

As input unit 111 and take-off equipment 112, can for example use the normal input/output device that is provided with on computers.Input unit 111 for example is keyboard or mouse.Take-off equipment 112 for example is display unit or stamping machine.Input unit 111 can be input file and/or output file with take-off equipment 112, maybe can be another computingmachine etc.

Storing device 130 is made up of the main storage means and the auxilary unit that are provided with on computers, and for example is used for being stored in various programs and the data that treatment unit 120 is carried out.Treatment unit 120 comprises the CPU of computingmachine, and moves through time variable control.

Treatment unit 120 comprises importing accepts handling part 121, colored graph picture and tumor region information acquisition portion 122, images match handling part (matching unit) 123, miRNA colored graph as tumor region extraction portion (designating unit) 124 and stained positive cell containing ratio calculating part (computing unit) 125.Colored graph picture and tumor region information acquisition portion 122 have above-mentioned image and obtain the two function of unit and information acquisition unit.

Handling part 121 is used to specify the pathological image that will diagnose from acceptance such as users through input unit 111 information and the information that is used to specify inspect-type is accepted in input.The example of inspect-type comprises the type of the cancer markers miRNA that will detect.And input is accepted handling part 121 and is stored those information in the diagnostic image information of storing device 130 in the object storage portion 131 with checking, and colored graph picture and tumor region information acquisition portion 122 are transferred in processing.Under the situation of this embodiment, being used to specify the information of the pathological image that will diagnose is image identifier 202.Image identifier 202 is HE colored graph picture or miRNA colored graph picture, and might specify the one or more of them.And being used to specify the information of inspect-type is about the painted project of miRNA, and might in cancer markers miRNA of the present invention, specify any one miRNA or two above miRNA.

The HE colored graph picture that colored graph picture and tumor region information acquisition portion 122 obtain to diagnose from dyeing image data base 113 and the information of miRNA colored graph picture and the tumor region HE colored graph picture; And with they be stored in respectively HE colored graph in the storing device 130 as data store 132, miRNA colored graph as data store 134 and HE colored graph as in the tumor region information storage part 133, and processing transferred to images match handling part 123.

Is under the painted situation of HE in diagnostic image information with the dyeing information 203 with image identifier 202 of checking storage in the object storage portion 131, and the view data 204 with image identifier 202 is stored in the HE colored graph as in the data store 132.And; Through inspect-type and sample neighbor information 205 with reference to storage in diagnostic image information and inspection object storage portion 131, the miRNA colored graph as data store 134 in the view data 204 of miRNA colored graph picture of the storage serial section sample adjacent with the HE image pattern that will diagnose.And the HE colored graph is stored in the HE colored graph as in the tumor region information storage part 133 as the information of tumor region information 206.

On the other hand; The dyeing information 203 with image identifier of storage is under the painted situation of miRNA in diagnostic image information and inspection object storage portion 131, the miRNA colored graph as data store 134 in storage have the view data 204 of image identifier 202.And, through reference sample neighbor information 205, the HE colored graph as data store 132 in the view data 204 of HE colored graph picture of the storage serial section sample adjacent with the miRNA that will diagnose dyeing image pattern.And, the HE colored graph as tumor region information storage part 133 in storage HE colored graph as the information of tumor region information 206.When specifying a plurality of image identifier, search, corresponding and store they each.

Images match handling part 123 reads HE colored graph picture and miRNA colored graph picture as data store 132 and miRNA colored graph as data store 134 from the HE colored graph respectively, and calculates the matched position of this HE colored graph picture and miRNA colored graph picture.And images match handling part 123 is stored matched position information in matched position information storage part 135, and processing is transferred to the miRNA colored graph as tumor region extraction portion 124.The example of matched position information comprises angle of rotation and horizontal/vertical misalignment width.Because HE colored graph picture and miRNA colored graph similarly are the image that obtains through the section of dyeing successive, so they maybe be very similar each other.For example, can use the color of same hue or different tones to carry out HE dyeing and miRNA dyeing.Yet, because HE colored graph picture and miRNA colored graph look like to carry out matching treatment, thus preferably, in miRNA dyeing, use with HE dyeing in the color of the different tones of color of use come staining cell.Usually, in HE dyeing, use phenodin that nucleus dyeing is blueness, and use eosin that tenuigenin is dyeed to pink.For example, can wait the tone that suitably is arranged on the color of using in the miRNA dyeing through the substance that show color or the fluorescent substance that will use.In images match, each image is by binaryzation, and, can use pure phase correlation method, sequential similar detection algorithm and use the method for distinguished point.

The miRNA colored graph reads HE colored graph as tumor region information, miRNA dyeing view data and matched position information as tumor region information storage part 133, miRNA colored graph as data store 134 and matched position information storage part 135 from the HE colored graph respectively as tumor region extraction portion 124; And, calculate the tumor region in miRNA dyeing view data.And, the miRNA colored graph as tumor region extraction portion 124 the miRNA colored graph as tumor region information storage part 136 in the information of tumor region in the storage miRNA dyeing view data, and processing transferred to stained positive cell containing ratio calculating part 125.

Stained positive cell containing ratio calculating part 125 reads miRNA dyeing view data and tumor region information as data store 134 and miRNA colored graph as tumor region information storage part 136 from the miRNA colored graph respectively.Count the quantity of the nuclear quantity of stained positive in tumor region and the negative cells nuclear that dyes then, and calculate stained positive cell containing ratio to export from take-off equipment 112.

With reference to the schema shown in the figure 3 to 6, will be described in the operation example of the system shown in Fig. 1 below as the embodiment 1B of diagnostic imaging support method according to the present invention and diagnostic imaging support program.In this embodiment, will describe based on following hypothesis: in miRNA dyeing, positive cell nuclear is colored and is blueness, and negative cells nuclear is colored and is red-brown.Yet, should be noted that to the invention is not restricted to this, and can be for example through being used to specify the general method of staining cell nuclear to count positive cell nuclear and negative cells nuclear.Specifically, for example,, staining cell nuclear can be come, and miRNA stained positive nucleus and miRNA dyeing negative cells nuclear can be counted through general dyeing process about carrying out the painted slide glass of miRNA.

In a word, the method for this embodiment is to be used to support the method based on the cancer diagnosis of pathological image, and comprises that following step (a) is to (e).And the program of this embodiment is the program that is used to support based on the diagnosis of the cancer of pathological image, and this program makes computingmachine carry out (a) to (e).

(a) image obtains step, obtains HE colored graph picture and miRNA colored graph picture as the pathological image that will diagnose;

(b) information acquisition step, the information of the tumor region of acquisition in HE colored graph picture;

(c) coupling step, the HE colored graph picture that obtains in the computed image acquisition step and the matched position of miRNA colored graph picture;

(d) given step is based on the information of the tumor region in HE colored graph picture that obtains in the information acquisition step and the information of the matched position that calculates in the step in coupling specifies in the tumor region in the miRNA colored graph picture; And

(e) calculation procedure, the information that is based on the tumor region in the specified miRNA colored graph of the given step picture is calculated the containing ratio of the stained positive cell in the tumor region of miRNA colored graph picture.

More detailed; When beginning to handle; Storage person under inspection identifier 201, image identifier 202, dyeing information 203, view data 204, sample neighbor information 205 and HE colored graph are as tumor region information 206 in dyeing image data base 113, and they all are a series of data of person under inspection.Tumor region information 206 is through calculating information that tumor region obtains in advance from HE colored graph picture or by the specified tumor region information of doctor.When in such state, starting treatment unit 120, beginning is in the processing shown in Fig. 3.

At first, accepting handling part 121 from input unit 111 to the input of treatment unit 120 provides: the image identifier of the image identifier of HE colored graph picture or miRNA colored graph picture, and it specifies diagnostic image; And, the requirement of miRNA inspection item, it specifies inspection object cancer markers miRNA.Input is accepted information that handling part 121 will be used to specify diagnostic image and the information that is used to specify the inspection object and is sent to colored graph picture and tumor region information acquisition portion 122 from the diagnostic image information of storing device 130 with inspection object storage portion 131.Then, handle colored graph picture and the tumor region information acquisition portion 122 (step S301) of transferring to.

Next, colored graph picture and tumor region information acquisition portion 122 in diagnostic image information and inspection object storage portion 131 to dyeing image data base 113 searching image identifiers.When the dyeing information 203 with specified image identifier is HE whens dyeing, colored graph picture and tumor region information acquisition portion 122 store the view data 204 with image identifier at the HE colored graph in as data store 132.And the HE colored graph is stored in the HE colored graph as in the tumor region information storage part 133 as tumor region information 206.And; Read in diagnostic image information and the miRNA inspection item of checking in the object storage portion 131; With reference to the sample neighbor information 205 in dyeing image data base 113; And, the miRNA colored graph as data store 134 in storage conduct and HE colored graph as the miRNA dyeing view data 204 of adjacent serial section sample.

On the other hand, when the dyeing information 203 with specified image identifier is that miRNA dyeing is, the miRNA colored graph as data store 134 in storage have the view data 204 of image identifier.And, with reference to the sample neighbor information 205 of dyeing image data base 113, and the HE colored graph as data store 132 in storage conduct and miRNA colored graph as the HE dyeing view data 204 of adjacent serial section sample.And, the HE colored graph as tumor region information storage part 133 in storage HE colored graph as tumor region information 206.Then, images match handling part 123 (step S302) is transferred in processing.

Images match handling part 123 calculate the HE colored graph as data store 132 in storage HE colored graph picture and the miRNA colored graph as data store 134 in the matched position of miRNA colored graph picture of storage.For example, behind the colour code of two images of adjustment, carry out the calculating of matched position through using pure phase correlation method.The matched position information that so obtains is stored in the matched position information storage part 135.The example of matched position information comprises angle of rotation and horizontal/vertical misalignment width.Then, processing is transferred to the miRNA colored graph as tumor region extraction portion 124 (step S303).

The miRNA colored graph as tumor region extraction portion 124 calculate from the HE colored graph as tumor region information storage part 133 the matched position information of tumor region information and storage in matched position information storage part 135 of HE colored graph picture of storage calculate the miRNA colored graph as data store 134 in tumor region in the miRNA colored graph picture of storage.The miRNA colored graph as tumor region extraction portion 124 the miRNA colored graph as tumor region information storage part 136 in the tumor region information of the miRNA colored graph picture that so obtains of storage.Then, stained positive cell containing ratio calculating part 125 (step S304) are transferred in processing.

Stained positive cell rate calculating part 125 be received in the miRNA colored graph as the miRNA dyeing view data of storage in the data store 134 and the miRNA colored graph as tumor region information storage part 136 in the tumor region information of storage.Then, stained positive cell rate calculating part 125 is counted the quantity of the nuclear quantity of stained positive in tumor region and the negative cells nuclear that dyes, and calculates stained positive cell containing ratio to export (step S305) from take-off equipment 112.When the stained positive nucleus is colored when being colored to blueness for red-brown and dyeing negative cells nuclear, counting be colored for brown nuclear quantity be colored nuclear quantity into blueness.For example, according in Fig. 4, the process shown in 5 and 6, and be described below, carry out this processing.

At first, cover the dye outside (step S401) of tumor region of view data of miRNA based on the miRNA that received dyeing view data and tumor region information.In tumor region, discern as the brown area of dying brown zone and the blue region (step S402) of dying blue zone through discriminatory analysis.

In this was handled, at first, view data was converted into hsv color space (step S501); Remove undyed zone (step S502) according to S (saturation ratio) and V (lightness), and, with the value scope of H (tone) from [0; 1] to [0.3,1.3] conversion (step S503).Whether H (tone) value of next, checking all pixels is included in any one of scope [0.3,0.8] and scope [0.8,1.3] (step S504).When in a zone, comprising all pixels, [0.3,0.8] is outputted as blue region, and [0.8,1.3] is outputted as brown area (step S507).When in two zones, having pixel, come calculated threshold t (step S505) through discriminatory analysis, and [0.3, t] is outputted as blue region, [t, 1.3] are outputted as brown area (step S506).

Next, in brown area, carry out nuclear and extract (step S403), and subsequently, in blue region, carry out nuclear and extract (step S404).In these steps, at first, when input brown area or blue region (step S601), stress the V ' value (step S602) that V (lightness) value obtains with distributing to calculate through use S type function according to the MV of V (lightness) value.Then, carry out the conversion to bianry image in the following manner: when V ' value was equal to or less than specific threshold, input area was set in the nuclear zone (=1); And when V ' was worth greater than threshold value, input area was set to (=0) (step S603) outside the nuclear zone.Next, carry out neighbor relatively, the position (step S604) of calculating nuclear through use Gaussian filter to bianry image.

Next, counting detected nuclear volume (step S405) in brown area, and counting detected nuclear volume (step S406) in blue region.At last, calculate the ratio of sum of quantity and the nuclear of brown nuclear, that is, and the quantity of brown nuclear/(quantity of the quantity of brown nuclear+blueness nuclear) (step S407).

The effect of this embodiment is described below.In this embodiment, image obtains the unit and obtains HE colored graph picture and miRNA colored graph picture, and the information acquisition unit obtains the information of the tumor region in HE colored graph picture.Thereafter, matching unit calculates the matched position of HE colored graph picture and miRNA colored graph picture.Subsequently, the designating unit information and the information of matched position that are based on the tumor region of HE colored graph picture specifies in the tumor region in the miRNA colored graph picture.Then, the computing unit information that is based on the tumor region in the miRNA colored graph picture is calculated the stained positive cell containing ratio in the tumor region of miRNA colored graph picture.Thus, can be used as the containing ratio that quantized value obtains the stained positive cell.As a result, become might the doctor can dye and carry out diagnosis through miRNA based on this quantized value.

And, because the case quantity of organizational diagnosis and cytodiagnosis in the last few years increases and pathologist's comparatively small amt, so the problem that forces the pathologist to work long hours has been arranged.In this, according to this embodiment, can alleviate doctor's etc. labor load.

In addition, according to this embodiment, through coupling as the HE colored graph picture and the miRNA colored graph picture of serial section sample image, can be with the tumor region of in HE colored graph picture, confirming and miRNA colored graph as corresponding.And, for example,, can discern brown area and blue region through H (tone) value is used discriminatory analysis.And, extract through in each brown area and blue region, carrying out nuclear, can calculate the ratio of sum of quantity and the nuclear of brown nuclear.Therefore,, helpful information might be provided, support diagnosis thus doctor's diagnosis through waiting the containing ratio that the stained positive cell is provided Xiang the doctor.

(embodiment 1C)

Fig. 7 is the block diagram that illustrates according to another example of image diagnosis support system of the present invention.The system of this embodiment be the different of system shown in Fig. 1 according to embodiment 1B: stained positive cell containing ratio calculating part 125 also calculates staining power except stained positive cell containing ratio.Unless otherwise indicated, those of other configurations and operation and embodiment 1B are identical.

In Fig. 7, the containing ratio of stained positive cell and staining power calculating part 725 read miRNA dyeing view data and tumor region information as data store 134 and miRNA colored graph as tumor region information storage part 136 from the miRNA colored graph respectively.Then; The quantity of stained positive cell containing ratio and the stained positive nuclear quantity of staining power calculating part 725 countings in tumor region and dyeing negative cells nuclear; With the containing ratio of calculating stained positive cell, and further calculate staining power to export from take-off equipment 112.

With reference at the schema shown in Fig. 8 to 10, be described in the operation example of the system shown in Fig. 7 below as the embodiment 1C of diagnostic imaging support method according to the present invention and diagnostic imaging support program.

The processing of this embodiment is with the different of processing of the embodiment 1B shown in Fig. 3: not only calculate the containing ratio of stained positive cell, and calculate staining power, and other operations are identical with in embodiment 1B those.

The containing ratio of stained positive cell and staining power calculating part 725 be received in the miRNA colored graph as the miRNA dyeing view data of storage in the data store 134 and the miRNA colored graph as tumor region information storage part 136 in the tumor region of storage.Then; The quantity of the containing ratio of stained positive cell and the stained positive nuclear quantity of staining power calculating part 725 countings in tumor region and dyeing negative cells nuclear; With the containing ratio of calculating stained positive cell, and calculate staining power (0: feminine gender, 1: the weak positive; 2: the medium positive, 3: strong positive) to export (step S805) from take-off equipment 112.For example, according in the process shown in Fig. 9 and 10, and be described below, carry out this processing.

This processing is all identical up to the step S407 of Fig. 9 with the processing at the embodiment 1B shown in Fig. 4.After step S407, in brown area, calculate nuclear staining intensity (step S908).

At first, import the brown area of in the step S402 of Fig. 9, confirming (step S1001).Then, according to MV and the distribution of V (lightness), calculate through using the S type function to stress the V ' (step S1002) that V obtains.When V ' value was equal to or less than specific threshold x, input area was set in the nuclear zone, and the quantity (step S1003) of counting pixel X wherein.

Next, the constant a, b and the c that satisfy 0＜a＜b＜c＜1 are set, and, confirm the to satisfy condition pixel quantity of V≤a and the ratio of the pixel quantity in the nuclear zone.When this ratio is not less than specific ratios (step S1004), it is outputted as staining power " 3: strong positive " (step S1005).If not this situation, then confirm the to satisfy condition pixel quantity of V≤b and the ratio of the pixel quantity in the nuclear zone.When this ratio is not less than specific ratios (step S1006), it is outputted as staining power " 2: the medium positive " (step S1007).If not this situation, then confirm the to satisfy condition pixel quantity of V≤c and the ratio of the pixel quantity in the nuclear zone.When this ratio is not less than specific ratios (step S1008), it is outputted as staining power " 1: the weak positive " (step S1009).If not this situation, then it is outputted as staining power " 0: feminine gender " (step S1010).

The effect of this embodiment is described below.Though the containing ratio of stained positive cell only is provided Xiang the doctor in embodiment 1B, in embodiment 1C, might wait Xiang the doctor not only provides the containing ratio of stained positive cell but also staining power is provided.Therefore, more helpful information might be provided to doctor's diagnosis, support diagnosis thus.Those of other effects of this embodiment and embodiment 1B are identical.

(embodiment 1D)

Figure 11 is the block diagram that illustrates according to another example of image diagnosis support system of the present invention.The system of this embodiment be the different of system shown in Fig. 1 according to embodiment 1B: the system of this embodiment has tumour and confirms and tumor region calculating part (tumor region computing unit) 1126.Unless otherwise indicated, other configurations are identical with in embodiment 1B those with operation.In addition, the miRNA colored graph picture of the sample of accumulation at least one HE colored graph picture, the conduct serial section adjacent and the sample neighbor information of HE colored graph picture and miRNA colored graph picture in dyeing image data base 113 with the sample of HE colored graph picture.In this embodiment, existing of the tumor region information of confirming from miRNA dyeing image calculation or by doctor etc. is dispensable.

In Figure 11; Colored graph picture and tumor region information acquisition portion 122 from dyeing image data base 113 obtain the HE colored graphs as 204, the miRNA colored graph as 204 and the HE colored graph as tumor region information 206, and the HE colored graph that they are stored in storing device 130 respectively as data store 132, miRNA colored graph as data store 134 and HE colored graph as in the tumor region information storage part 133.At this, when having tumor region information 206, handle and transfer to images match handling part 123; Yet, when not having tumor region information 206, handle and transfer to the definite and tumor region calculating part 1126 of tumour.

Tumour is confirmed to read HE dyeing view data from the HE colored graph as data store 132 with tumor region calculating part 1126, confirms tumour and calculates tumor region, and images match handling part 123 is transferred in processing.As method and the tumor region Calculation Method that tumour is confirmed, for example, can use those disclosed in patent document 1.

The effect of this embodiment is described below.In this embodiment,, can confirm to carry out a series of processing through tumour is provided with the tumor region calculating part even when in HE colored graph picture, not confirming tumour.Therefore, can wait the diagnostic message that the integration from cancer diagnosis to the immunohistochemical staining diagnostic imaging is provided Xiang the doctor.Therefore, might provide the information that benefits diagnosis, support diagnosis thus.Other effects of this embodiment are identical with in embodiment 1B those.

Simultaneously, in this embodiment, calculate staining power with the containing ratio of stained positive cell can resembling in embodiment 1C.

(embodiment 1E)

Figure 12 is the block diagram that illustrates according to another example of image diagnosis support system of the present invention.The system of this embodiment comprises: slide glass imaging portion 1222 (slide glass acquisition unit), and it is substituted in colored graph picture and the tumor region information acquisition portion 122 shown in Fig. 1 etc.; Slide glass DB 1213, its substituted dyeing image data base 113; And, diagnosis slide glass information and inspection object storage portion 1231, it replaces diagnostic image information and inspection object storage portion 131.And system has slide glass imaging device 1214 and tumour is confirmed and tumor region calculating part 1126.Unless otherwise indicated, other configurations are identical with in embodiment 1B those with operation.

The miRNA dyeing slide glass of the sample of the serial section that at least one HE of accumulation dyeing slide glass, conduct are adjacent with the sample of HE dyeing slide glass in slide glass DB 1213 and the sample neighbor information of HE dyeing slide glass and miRNA dyeing slide glass.Relevant information about the person under inspection is associated with each slide glass through person under inspection's identifier.The specified slide glass of slide glass imaging device 1214 imagings is to convert numerical data into.

Input is accepted handling part 121 and is used to specify the information (slide identifier symbol) of the slide glass that will diagnose and is used to specify the information of inspect-type from acceptance such as users through input unit 111.Then, input is accepted handling part 121 they is stored in the diagnosis slide glass information and inspection object storage portion 1231 of storing device 130, and slide glass imaging portion 1222 is transferred in processing.

Slide glass imaging portion 1222 obtains HE dyeing slide glass and miRNA dyeing slide glass as the adjacent sample that will diagnose from slide glass DB 1213.And the slide glass that slide glass imaging portion 1222 is obtained by slide glass imaging device 1214 through imaging and convert them to numerical data obtains HE colored graph picture and miRNA image.Then, the HE colored graph that these images are respectively stored in storing device 130, and is handled and is transferred to tumour and confirm and tumor region calculating part 1126 as in the data store 134 as data store 132 and miRNA colored graph.As stated, in this embodiment, slide glass imaging portion 1222 has the function that slide glass obtains unit and image acquisition unit.

The effect of this embodiment is described below.In this embodiment, even when the pathology slide glass is not converted into numerical data, also through providing slide glass imaging device, slide glass data and slide glass imaging portion to carry out a series of processing.Therefore, can wait to provide Xiang doctor and form images to cancer diagnosis and further to the comprehensive diagnos information of immunohistochemistry ground dyeing diagnostic imaging from slide glass.Therefore, might provide the information that benefits diagnosis, support diagnosis thus.Other effects of this embodiment are identical with in embodiment 1B those.

The invention is not restricted to above-mentioned exemplary, and can carry out various modifications.For example, above-mentioned exemplary relates to input and accepts the situation that handling part 121 accepts to be used to specify the image identifier of diagnostic image.Yet input is accepted handling part 121 and can be accepted to diagnose person under inspection's identifier of target to come the alternative image identifier.In this case, the image with person under inspection's identifier and tumor region information can be searched for to dyeing image data base 113 by colored graph picture and tumor region information acquisition portion 122.

Second embodiment

Comprising according to cancer pathology diagnostic imaging support equipment of the present invention (below be called " diagnostic imaging support equipment "):

The mode of learning input block, being used for obtaining with the tumour from the pathological image that will be used to learn is the image at center, and said image is imported to it as mode of learning;

The mode of learning storage unit is used to store and keep the said mode of learning with classification information;

The candidate feature generation unit is used to produce a plurality of candidate feature;

Characteristic is confirmed the unit, is used to use the characteristic set of the characteristic that the candidate feature that is produced by said candidate feature generation unit confirms to be suitable for diagnosing;

The characteristic storage unit is used to store and keeps confirming the characteristic set that the unit is confirmed by said characteristic;

The sorted table generation unit is used to produce sorted table;

The pattern input block, being used for obtaining with candidate's tumour from the pathological image that will diagnose is the image at center, and said image is imported as input pattern;

Feature extraction unit is used for extracting characteristic from said input pattern; And

Diagnosis unit is used for diagnosing based on said characteristic.Said characteristic confirms that the unit calculates the characteristic of each the said mode of learning corresponding with each said candidate feature; And first characteristic as said characteristic set is confirmed candidate feature; Candidate feature is hereto measured peak about the mutual information of the classification information of said mode of learning set; And; Under the condition of known definite characteristic, confirm candidate feature in regular turn as the next characteristic of said feature set; Candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a mode of learning being associated is measured peak.Said list generation unit uses each characteristic of incompatible each the said mode of learning of calculating of said feature set; Generation comprises the said sorted table of each characteristic and the said classification information of said mode of learning, and uses the said sorted table said pattern of classifying.Said feature extraction unit is used each characteristic of the incompatible calculating input pattern of said feature set.Said diagnosis unit is diagnosed said input pattern according to the result and the said sorted table of said diagnosis.

Diagnostic imaging support equipment according to the present invention comprises:

The mode of learning input block, being used for obtaining with the tumour from the pathological image that will be used for learning is the image at center, and said image is imported to it as mode of learning;

The characteristic storage unit, the set that is used to store and keep confirming the characteristic that the unit is confirmed by said characteristic;

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said characteristic is confirmed the said mode of learning set of the pre-determined quantity that the unit is prepared according to the value of said characteristic to move; Calculate the characteristic of each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature hereto; Mutual information about the classification information of said mode of learning set is measured peak; Said characteristic according to so confirming is distributed the said mode of learning with weight, causes the migration of said mode of learning to one of said set corresponding with said characteristic in regular turn, and; Under the condition of known information about the set that comprises said mode of learning and said definite characteristic respectively; Confirm candidate feature in regular turn as the next characteristic of said characteristic set, candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak.Said sorted table generation unit uses each characteristic of incompatible each the said mode of learning of calculating of said feature set; Generation comprises the said sorted table of each characteristic and the said classification information of said mode of learning, and uses the said sorted table said pattern of classifying.Said feature extraction unit is used each characteristic of the incompatible calculating input pattern of said feature set.The migration that said diagnosis unit causes each said input pattern according to each characteristic and the migration table of said input pattern; Said conversion table has write down the set under the said mode of learning when confirming each characteristic of said characteristic set in regular turn, and said diagnosis unit is according to diagnosing said input pattern as the set under the result's of said migration the said input pattern.

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said characteristic is confirmed the said mode of learning set of the pre-determined quantity that the unit is prepared according to the value of said characteristic to move; Calculate the characteristic of each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature hereto; The mutual information of classification information about the said mode of learning of a set is measured peak; Said characteristic according to so confirming is distributed the said mode of learning with weight, causes the migration of said mode of learning to one of said set corresponding with said characteristic in regular turn, and; Under the condition of known information about the set that comprises said mode of learning and said definite characteristic respectively; Confirm the candidate feature of levying in regular turn as the next one of said characteristic set, candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak.Said sorted table generation unit uses each characteristic of incompatible each the said mode of learning of calculating of said feature set, and uses the sorted table of each characteristic comprise said mode of learning and the classifiedinformation said pattern of classifying.Said feature extraction unit is used each characteristic of the incompatible calculating input pattern of said feature set, and said each characteristic is used in reference to the said characteristic that is shown in certain order and gets the probability of preset value.Said diagnosis unit calculates the probability that said input pattern comprises predetermined classification information according to each characteristic and the migration table of said input pattern; Carry out said diagnosis then, said conversion table has write down the set under the said mode of learning when confirming each characteristic of said characteristic set in regular turn.

In said diagnostic imaging support equipment of the present invention; Preferably; The pixel in said mode of learning input block and the said pattern input block look zone under the R of each pixel prestained said pathological image, G and B value select to belong to the nucleus of predetermined tumour; Calculating is in the distribution center in said look zone and belong to the distance between each pixel in said look zone; To each pixel distributing signal, detect the peak value of the signal distributions in said pathological image according to said distance, and input is that the image at center is as said mode of learning with said peak value.

In said diagnostic imaging support equipment of the present invention, preferably, the candidate feature that is produced by said characteristic generation unit comprises the candidate feature that obtains from the feature extraction function.

In said diagnostic imaging support equipment of the present invention, preferably, the said candidate feature that is produced by said characteristic generation unit comprises that use becomes the candidate feature that feature extraction function that the Gabor function obtains obtains again through normalization method.

In diagnostic imaging support equipment according to the present invention, preferably, the candidate feature that is produced by said candidate feature generation unit comprises the candidate feature of the color that is used to distinguish tumour.

In diagnostic imaging support equipment according to the present invention, preferably, the signal and the predetermined threshold of each pixel that comprises in the said mode of learning that said characteristic is confirmed will be calculated by said mode of learning input block the unit are made comparisons.

In diagnostic imaging support equipment according to the present invention; The MV of the signal of each pixel that comprises in the mode of learning that preferably, said characteristic is confirmed will be calculated by said mode of learning input block the unit and near the signal of the pixel said pixel is made comparisons.

In diagnostic imaging support equipment according to the present invention, preferably, said characteristic confirms that the unit uses the predetermined noise parameter of each said candidate feature to come said mode of learning each is operated.

In diagnostic imaging support equipment according to the present invention, preferably, said characteristic confirms that the characteristic that the unit calculates said mode of learning gets the probability of preset value, as the characteristic of each the said mode of learning corresponding with each said candidate feature.

In diagnostic imaging support equipment according to the present invention, preferably, in the time can irrespectively classifying said mode of learning with the value of characteristic, said sorted table generation unit replaces the value of characteristic with redundancy at the relevant position place of said sorted table.

In diagnostic imaging support equipment according to the present invention, preferably, each characteristic of said input pattern is said characteristic is got the probability of preset value when certain order a value; And said diagnosis unit is judged through following manner: each the said feature mode that uses said characteristic to calculate in said sorted table, to comprise is got the probability of the preset value of classifiedinformation.

And diagnostic imaging support program according to the present invention is characterised in that it can be so that computingmachine be carried out image diagnosing method of the present invention.Diagnostic imaging support program of the present invention is the diagnostic imaging support program that for example is used for diagnostic imaging support equipment, and said diagnostic imaging support equipment comprises:

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said program makes units corresponding carry out following processing: handle; Wherein said characteristic confirms that the unit calculates the characteristic of each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature is hereto measured peak about the mutual information of the classification information of said mode of learning set; And; Under the condition of known said definite characteristic, confirm candidate feature in regular turn as the next characteristic of said characteristic set; Candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning that is associated is measured peak;

Handle, wherein, said sorted table generation unit uses each characteristic of incompatible each the said mode of learning of calculating of said feature set, and uses the said sorted table of each characteristic that comprises said mode of learning and the said classification information said pattern of classifying;

Handle, wherein, said feature extraction unit is used each characteristic of the incompatible calculating input pattern of said feature set; And

Handle, wherein, said diagnosis unit is diagnosed said input pattern according to said result calculated and said sorted table.Specifically, program according to the present invention is to make the corresponding units of said diagnostic imaging support equipment carry out the program of above-mentioned treatment step.

Diagnostic imaging support program of the present invention is the diagnostic imaging support program that is used for a kind of diagnostic imaging support equipment, and said diagnostic imaging support equipment comprises:

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said program makes corresponding units carry out following processing: handle; Wherein said characteristic is confirmed the set of the said mode of learning of the pre-determined quantity that the unit is prepared according to the value of said characteristic to move; Calculate the characteristic of each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature is hereto measured peak about the mutual information of the classification information of said mode of learning set, distributes the said mode of learning with weight according to the said characteristic of so confirming; Cause of the migration of said mode of learning in regular turn to one of said set corresponding with said characteristic; And, under the condition of known information about the set that comprises said mode of learning and said definite characteristic respectively, confirm candidate feature in regular turn as the next characteristic of said characteristic set; Candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak;

Handle; Wherein, Said sorted table generation unit uses each characteristic of incompatible each the said mode of learning of calculating of said feature set, produces the said sorted table of each characteristic and the said classification information comprise said mode of learning, and uses the said sorted table said pattern of classifying; And

Handle; Wherein, The migration that said diagnosis unit causes each said input pattern according to each characteristic and the migration table of said input pattern; Said migration table has write down the set under the said mode of learning when confirming each characteristic of said characteristic set in regular turn, and said diagnosis unit is according to diagnosing said input pattern as the set under the result's of said migration the said input pattern.Specifically, program according to the present invention is to be used to make the corresponding units of said diagnostic imaging support equipment to carry out the program of above-mentioned treatment step.

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said program makes corresponding units carry out a series of processing described below: handle; Wherein, Said characteristic is confirmed the said mode of learning set of the pre-determined quantity that the unit is prepared according to the value of said characteristic to move; Calculate the characteristic of each the said mode of learning corresponding, confirm candidate feature, hereto candidate feature as first characteristic of said characteristic set with each said candidate feature; Mutual information about the classification information of said mode of learning set is measured peak; Said characteristic according to so confirming is distributed the said mode of learning with weight, causes the migration of said mode of learning to one of said set corresponding with said characteristic in regular turn, and; Under the condition of known information about the set that comprises said mode of learning and said definite characteristic respectively; Confirm candidate feature in regular turn as the next characteristic of said characteristic set, candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak;

Handle, wherein, said sorted table generation unit uses each characteristic of incompatible each the said mode of learning of calculating of said feature set, and uses the sorted table of each characteristic that comprises said mode of learning and the classifiedinformation said pattern of classifying;

Handle, wherein, said feature extraction unit is used each characteristic of the incompatible calculating input pattern of said feature set, and said each characteristic is used to indicate said characteristic when certain order, to get the probability of preset value; And

Handle; Wherein, Said diagnosis unit calculates the probability that said input pattern comprises predetermined classification information according to each characteristic and the migration table of said input pattern; Carry out said diagnosis then, said migration table has write down the set under the said mode of learning when confirming each characteristic of said characteristic set in regular turn.Specifically, program according to the present invention is the program that is used to make the corresponding units of said diagnostic imaging support equipment carry out above-mentioned treatment step (below identical).

In said diagnostic imaging support program of the present invention; Preferably; For example; Said mode of learning input block and said pattern input block comprise: be used for the processing of the pixel in the look zone under the nucleus of R, G and the B value of each pixel of prestained said pathological image, selecting to belong to predetermined treatment, be used to calculate in the distribution center in said look zone and belong to the processing of the distance between each pixel in said look zone; Be used for according to of the processing of said distance to each pixel distributing signal; Be used for detecting processing at the peak value of the distribution of the signal of said pathological image; And being used to import with said peak value is the processing that the image at center is used as said mode of learning.

In said diagnostic imaging support program of the present invention, preferably, the candidate feature that is produced by said characteristic generation unit comprises the candidate feature that obtains from the feature extraction function.

In said diagnostic imaging support program of the present invention, preferably, the said candidate feature that is produced by said characteristic generation unit comprises that use becomes the candidate feature that feature extraction function that the Gabor function obtains is obtained again through normalization method.

In said diagnostic imaging support program of the present invention, preferably, the candidate feature that is produced by said candidate feature generation unit comprises the candidate feature of the color that is used to distinguish tumour.

Said diagnostic imaging support program of the present invention for example preferably includes: handle, wherein, the signal and the predetermined threshold of each pixel that comprises in the said mode of learning that said characteristic is confirmed will be calculated by said mode of learning input block the unit are made comparisons.

Said diagnostic imaging support program of the present invention for example preferably includes: handle; The MV of the signal of each pixel that comprises in the mode of learning that wherein, said characteristic is confirmed will be calculated by said mode of learning input block the unit and near the signal of the pixel said pixel is made comparisons.

Said diagnostic imaging support program of the present invention for example preferably includes: handle, wherein, said characteristic confirms that the unit uses the predetermined noise parameter of each said candidate feature to operate for said mode of learning each.

Said diagnostic imaging support program of the present invention for example preferably includes: handle; Wherein, Said characteristic confirms that the characteristic that the unit calculates said mode of learning gets the probability of preset value, as the characteristic of each the said mode of learning corresponding with each said candidate feature.

Said diagnostic imaging support program of the present invention for example preferably includes: handle; Wherein, In the time can irrespectively classifying said mode of learning with the value of characteristic, said sorted table generation unit replaces the value of characteristic with redundancy in the position that is associated of said sorted table.

In diagnostic imaging support program according to the present invention, preferably, each of the characteristic of said input pattern is said characteristic is got the probability of preset value when certain order a value; And said program further comprises processing, and wherein, said diagnosis unit is got the probability of the preset value of classifiedinformation through each of the said feature mode that uses said characteristic to calculate in said sorted table, to comprise, so that confirm.

The invention provides a kind of pathological diagnosis support method of using the pathological diagnosis supportive device, said pathological diagnosis supportive device comprises:

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said characteristic confirms that the unit carries out following step: the characteristic of calculating each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature hereto; Mutual information about the classification information of said mode of learning set is measured peak; And under the condition of known said definite characteristic, confirm candidate feature in regular turn as the next characteristic of said characteristic set; Candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak.Said sorted table generation unit is carried out following step: use each characteristic of incompatible each the said mode of learning of calculating of said feature set, and use the said sorted table of each characteristic comprise said mode of learning and the said classification information said pattern of classifying.Said feature extraction unit is carried out the step of each characteristic of using the incompatible calculating input pattern of said feature set.The step of said input pattern is diagnosed in said diagnosis unit execution according to said result calculated and said sorted table.In the method for the invention, the use of diagnostic imaging support equipment is dispensable, and it can be a method of wherein carrying out above-mentioned corresponding each step.

The invention provides a kind of image pathological diagnosis support method of using the pathological diagnosis supportive device, said pathological diagnosis supportive device comprises:

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said characteristic confirms that the unit carries out following steps: the set of the said mode of learning of the pre-determined quantity of preparing according to the value of said characteristic to move; Calculate the characteristic of each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature is hereto measured peak about the mutual information of the classification information of said mode of learning set, distributes the said mode of learning with weight according to the said characteristic of so confirming; Cause of the migration of said mode of learning in regular turn to one of said set corresponding with said characteristic; And, under the condition of known said definite characteristic, confirm candidate feature in regular turn as the next characteristic of said characteristic set; Candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak.Said sorted table generation unit is carried out following steps: use each characteristic of incompatible each the said mode of learning of calculating of said feature set, and use the said sorted table of each characteristic that comprises said mode of learning and the said classification information said pattern of classifying.Said feature extraction unit is carried out the step of each characteristic of using the incompatible calculating input pattern of said feature set.Said diagnosis unit is carried out following steps: the migration that causes each said input pattern according to each characteristic and the migration table of said input pattern; Said migration table has write down the set under the said mode of learning when confirming each characteristic of said characteristic set in regular turn, and according to diagnosing said input pattern as the set under the result's of said migration the said input pattern.In the method for the invention, the use of diagnostic imaging support equipment is dispensable, and it can be a method of wherein carrying out above-mentioned corresponding each step.

The sorted table generation unit is used to produce sorted table;

Diagnosis unit is used for diagnosing based on said characteristic.Said characteristic confirms that the unit carries out following steps: the set of the said mode of learning of the pre-determined quantity of preparing according to the value of said characteristic to move; Calculate the characteristic of each the said mode of learning corresponding with each said candidate feature; And confirm candidate feature as first characteristic of said characteristic set; Candidate feature is hereto measured peak about the mutual information of the classification information of said mode of learning set, distributes the said mode of learning with weight according to the said characteristic of so confirming; Cause of the migration of said mode of learning in regular turn to one of said set corresponding with said characteristic; And, under the condition of known said definite characteristic, confirm candidate feature in regular turn as the next characteristic of said characteristic set; Candidate feature hereto, the mutual information between the said classification information of the characteristic of each mode of learning corresponding with each candidate feature and a said mode of learning being associated is measured peak.Said sorted table generation unit is carried out following steps: use each characteristic of incompatible each the said mode of learning of calculating of said feature set, and use the sorted table of each characteristic that comprises said mode of learning and the classifiedinformation said pattern of classifying.Said feature extraction unit is carried out following steps: use each characteristic of the incompatible calculating input pattern of said feature set, said each characteristic is used to each characteristic of indicating said characteristic when certain order, to get the probability of preset value.Said diagnosis unit is carried out following steps: each characteristic and migration table according to said input pattern calculate the probability that said input pattern comprises predetermined classification information; Carry out said diagnosis then, said migration table has write down the set under the said mode of learning when confirming each characteristic of said characteristic set in regular turn.In the method for the invention, the use of diagnostic imaging support equipment is dispensable, and it can be a method of wherein carrying out above-mentioned corresponding each step.

In said diagnostic imaging support method of the present invention; Preferably, said mode of learning input block and said pattern input block are carried out: the step of the pixel that the look under R, G and the B value of each pixel prestained said pathological image select to belong to the nucleus of predetermined tumour is regional; Calculating is in the distribution center in said look zone and belong to the step of the distance between each pixel in said look zone; According to the step of said distance to each pixel distributing signal; The step of the peak value of the distribution of the signal of detection in said pathological image; With input be the step that the image at center is used as said mode of learning with said peak value.

In said diagnostic imaging support method of the present invention, preferably, the candidate feature that is produced by said characteristic generation unit comprises the candidate feature that obtains from the feature extraction function.

In said diagnostic imaging support method of the present invention, preferably, the said candidate feature that is produced by said characteristic generator unit comprises that use becomes the candidate feature that feature extraction function that the Gabor function obtains obtains again through normalization method.

In said diagnostic imaging support method of the present invention, preferably, the candidate feature that is produced by said candidate feature generation unit comprises the candidate feature of the color that is used to distinguish tumour.

In said diagnostic imaging support method of the present invention, preferably, said characteristic is confirmed unit execution following steps: the signal and the predetermined threshold of each pixel that will in the said mode of learning of being calculated by said mode of learning input block, comprise are made comparisons.

In said diagnostic imaging support method of the present invention; Preferably, said characteristic is confirmed unit execution following steps: the MV of the signal of each pixel that will in the mode of learning of being calculated by said mode of learning input block, comprise and near the signal of the pixel said pixel is made comparisons.

In said diagnostic imaging support method of the present invention, preferably, said characteristic is confirmed unit execution following steps: use the predetermined noise parameter of each said candidate feature to operate for each said mode of learning.

In said diagnostic imaging support method of the present invention; Preferably; Said characteristic confirms that the unit carries out following steps: the characteristic of calculating said mode of learning is got the probability of preset value, as each the said mode of learning corresponding with each said candidate feature characteristic.

In said diagnostic imaging support method of the present invention; Preferably; In the time can irrespectively classifying said mode of learning with the value of characteristic, said sorted table generation unit is carried out following steps: the value that replaces characteristic in the position that is associated of said sorted table with redundancy.

In said diagnostic imaging support method of the present invention, preferably, each characteristic of said input pattern is said characteristic is got the probability of preset value when certain order a value; And said diagnosis unit is carried out the step of judging through following manner: each the said feature mode through using said characteristic to calculate in said sorted table, to comprise is got the probability of the preset value of classifiedinformation.

The invention provides a kind of image diagnosis support system, comprising:

The information processing terminal is used to preserve the pathological image data, said pathological image data comprise that pathological image and said image have for the unique information of patient; And

The diagnostic imaging server is used to diagnose said pathological image data.

Said diagnostic imaging server comprises:

According to diagnostic imaging support equipment of the present invention, be used for diagnosing the said pathological image that comprises in said pathological image data; And

The diagnostic result storage unit is used to store from the diagnostic result of said diagnostic imaging support equipment and for the unique said information of said patient.Said information processing terminal request diagnostic result is with sending for the unique information of patient; And the diagnostic imaging server will be made comparisons for the unique information of patient with what store together with diagnostic result for the unique information of patient from what the information processing terminal received; Then; If be complementary for the unique information of patient with what store for the unique information of patient from what the information processing terminal received, then send diagnostic result to the said information processing terminal together with diagnostic result.

Image diagnosis support system of the present invention preferably further comprises accounting server, is used for the amount of the cost of use of the difference memory image diagnosis supportive device and the information processing terminal.

In image diagnosis support system of the present invention, preferably, when diagnostic result cell stores diagnostic result, the amount of the cost of use of accounting server accumulative image diagnostic support system.

In image diagnosis support system of the present invention, preferably, when the information processing terminal receives diagnostic result, the amount of the cost of use of accounting server cumulative information processing terminal.

Server according to the present invention is the server that in image diagnosis support system of the present invention, uses.This server comprises: the server side transmitting element, its in server via network of communication to terminal transmits information; The server side acceptor unit, it receives the information of sending via network of communication from the terminal; According to diagnostic imaging support equipment of the present invention, be used to use the pathological image data to diagnose the person under inspection; And the diagnostic result storage unit is used for the diagnostic result of memory image diagnosis supportive device and for the unique information of patient.Server will be made comparisons for the unique information of patient with what store together with diagnostic result for the unique information of patient from what the information processing terminal received; If be complementary for the unique information of patient with what store for the unique information of patient from what the information processing terminal received then, then send diagnostic result to the said information processing terminal together with diagnostic result.

Terminal of the present invention is the terminal of in image diagnosis support system of the present invention, using.This terminal is the information processing terminal, is used to preserve the pathological image data for the unique information of patient that comprise that pathological image and image have.This terminal comprises: the end side transmitting element, and it sends via network of communication the information in the terminal to server; And, the end side receiving element, it receives the information of sending from server via network of communication.This terminal request is sent and is had the diagnostic result for the unique information of patient, and receives the diagnostic result that sends from server.

According to diagnostic imaging support equipment of the present invention, diagnostic imaging support method, diagnostic imaging support program and image diagnosis support system; When the change that in considering nucleus and its peripheral organization etc., occurs is the meaning in optimum or pernicious in the character of differentiation tumour; Extract the subimage (view data of subimage) that mainly comprises a nucleus and a matter from pathological image; And subimage is stored as mode of learning and input pattern.Based on subimage, can in the short period of time, confirm the existence of tumour accurately or not exist with it optimum or pernicious.In the present invention, pathological image is above-mentioned cancer markers colored graph picture.

According to aforesaid second embodiment of the present invention; For example; When the character that tumour is being differentiated in the change that in considering nucleus and its peripheral organization etc., occurs is the meaning in optimum or pernicious, the extraction subimage, and subimage is stored as mode of learning and input pattern.Therefore, based on subimage, can be in the short period of time confirm the existence of tumour for example accurately or do not exist with it optimum or pernicious.

Below, reference implementation scheme 2A and 2B describe the present invention.Below, the cancer markers colored graph looks like to be called as " miRNA colored graph picture ".Yet, should be noted that the present invention never is limited to these embodiments.

(embodiment 2A)

Figure 13 is the block diagram that illustrates according to the configuration of the diagnostic imaging support equipment of this embodiment.As shown in Figure 13, comprise that according to the diagnostic imaging support equipment of this embodiment mode of learning input block 1300, mode of learning storage unit 1301, candidate feature generation unit 1302, characteristic confirm unit 1303, characteristic storage unit 1304, sorted table generation unit 1305 and sorted table 1306.

Mode of learning input block 1300 extracts the subimage that comprises nucleus and tenuigenin etc. from miRNA colored graph picture, in mode of learning storage unit 1301, stores subimage then.

Mode of learning storage unit 1301 is unit of storing therein and the subimage of preserving the desired amt that is used to learn.

Candidate feature generation unit 1302 is to use the set of characteristic parameters of pre-determined quantity to come to generate in regular turn the unit of candidate feature.

Characteristic is confirmed unit 1303 definite best features set that is suitable for pattern discrimination most in the candidate feature that is produced by candidate feature generation unit 1302.

Characteristic storage unit 1304 is the characteristic set that unit 1303 is confirmed is confirmed in storing therein and preservation by characteristic unit.

Sorted table generation unit 1305 is to use by characteristic confirms that unit 1303 definite characteristic sets generate the unit of diagnostic sorted table 1306.

Referring to Figure 14, characterization is confirmed the process of processing.Figure 14 is the schema that the definite process of in the diagnostic imaging support equipment according to this embodiment, carrying out of handling of characteristic is described.

Candidate feature generation unit 1302 comes to generate in regular turn candidate feature (S1401) according to preassigned a large amount of (for example, N) set of characteristic parameters.In this embodiment; In N set of characteristic parameters, parameter sets 1 to N_1 is the candidate feature about texture, and parameter sets N_1+1 to N_1+N_2 is the candidate feature about color; And parameter sets N_1+N_2+1 to N is about using the candidate feature of the average color of peripheral pixel.Though in this embodiment, specified the candidate feature about the average color of the peripheral pixel of texture, color and use, candidate feature is not limited thereto.That is, confirm that the required and any element that in pathological image, comprise of the characteristic of each pixel can be used as candidate feature and produces.

Next, explanation is confirmed the method by the characteristic of subimage processing according to the candidate feature that is produced by candidate feature generation unit 1302.Confirm characteristic through following procedure 1 to 3 any.

Program 1

Candidate feature generation unit 1302 at first obtains s set of characteristic parameters.S=1 to N, and processing begins with s=1.If s≤N_1, then s set of characteristic parameters of candidate feature generation unit 1302 usefulness (k_s, r ₀_ s, σ _ s th_s) replaces (k, r ₀, σ th), comprises parameter k, r with generation ₀Multiple change Gabor function G ab and Gaussian function G shown in following expression (1) with σ.If s≤N_1, the subimage of the color that then will in mode of learning storage unit 1301, store as mode of learning is a gray level image according to grey level transition.Use the gray level image that so obtains to come calculated characteristics c.

Gab(r；k，r ₀、σ)

＝exp(ik(r-r ₀)-|r-r ₀| ²/(2σ ²))

G(r；r ₀，σ) …(1)

＝exp(-|r-r ₀| ²/(2σ ²)/(2πσ ²))

In expression formula (1), r=(x, y) indicating positions vector, and i ²=-1.Candidate feature generation unit 1302 confirms that to characteristic unit 1303 is provided at multiple change Gabor function G ab and the threshold parameter th and the identification number s (S1402) of Gaussian function G and candidate feature in the expression formula (1).

It is many to data item that mode of learning storage unit 1301 confirms that to characteristic unit 1303 sends, and wherein each comprises predetermined M number of sub images I_t (r, i_rgb) the classification qt (t=1 to M) (step S1403) under one of (t=1 to M) and this subimage.In the explanation of this embodiment, simple for what explain, this device uses two classifications (q=0 or 1).The invention is not restricted to this, and be applicable to three classifications or the more situation of comprising certainly.

The subimage that use receives from mode of learning storage unit 1301 in regular turn, characteristic confirm that unit 1303 uses candidate feature to come calculated characteristics c (step S1404) according to following expression (2).Candidate feature is multiple change Gabor function and Gaussian function and other parameter of for example expression formula (1).Suppose that t mode of learning is that (r i_rgb), repeatedly carries out calculating for all modes of learning (M pattern) to I_t.

a = {| \underset{r}{Σ} I_t (r, i_rgb) Gab (r; k, r_{0}, σ) |}^{2}

/ \underset{r}{Σ} I_t {(r, i_rgb)}^{2} G (r; r_{0}, σ) \cdot \cdot \cdot (2)

If c=1 is a >=th

C=0 otherwise

In expression formula (2) up, denominator is normalization method (stdn) factor, is used for suppressing because the variation of " a " value that the size of mode of learning (brightness of image) causes.This denominator can be replaced with another normalizing factor.According to the mode of learning that will handle, can omit this normalizing factor.

Program 2

Candidate feature generation unit 1302 at first obtains s set of characteristic parameters.S is 1 to N, and handles and begin with s=1.If N_1+1≤s≤N_1+N_2, then candidate feature generation unit 1302 is confirmed s set of characteristic parameters to be provided unit 1303 to characteristic (x_s, y_s is color_index) with the identification number s (step S1402) of this candidate feature.

In this manual, (x_s, y_s) indication is used for confirming the locations of pixels at the characteristic c of subimage, and color_inde (=1 to 4) the expression color corresponding with characteristic c.

Incidentally, color_index (=1 to 4) corresponding to for example color, is described below.

The miRNA dyeing of the cell of cancer markers miRNA is the important factor that characterizes in the pathological image.Therefore, preferably, use distinct colors to come staining cell nuclear, tenuigenin, a matter and hole etc.For the miRNA colored graph picture that produces from such dyeing, color_index=1,2,3,4 distribute to the color in nucleus, tenuigenin, a matter and hole respectively.

Mode of learning storage unit 1301 confirms that to characteristic unit 1303 provides data item right, wherein each comprise a predetermined M mode of learning (color sub image) I_t (r, one of i_rgb) with this mode of learning under classification qt (t=1 to M) (step SA1403).R is a pixel coordinate, and i_rgb=1 to 3 is parameters of the r that is used for specified pixel, g, b signal, and t=1 to M.In this embodiment, simple for what explain, two classifications of this apparatus processes (q=0 or 1).The invention is not restricted to this, and be applicable to three classifications or the more situation of comprising certainly.

For the mode of learning (step S1403) that receives from mode of learning storage unit 1301, characteristic confirms that unit 1303 confirms to be arranged in the position of mode of learning (x_s, the color of pixel of y_s) locating in the following manner.If this color with by the specified color-match of parameter c olor_index, then characteristic confirms that the value of unit 1303 characteristic c is set to 1.Otherwise characteristic confirms that unit 1303 these values are set to 0 (step S1404).Suppose that t mode of learning is that (r i_rgb), then repeatedly is used for the operation of the value of definite characteristic c to I_t for all modes of learning (M pattern).

For nuclear color (color_index=1), the phenodin signal and the threshold value (for example, 0.25) of each pixel of in subimage detects, calculating are made comparisons.If this signal, confirms then that color of pixel is nuclear color greater than threshold value.In order to confirm threshold value adaptively, then only near the pixel of being considered, the phenodin signal value is at least 0.25 pixel for example, addition phenodin signal value each other calculates the MV of phenodin signal value thus.This MV for example multiply by 0.9, with the threshold value of the pixel that obtains to be considered.If the phenodin signal value of the pixel of being considered greater than threshold value, then can be regarded color of pixel as nuclear color.

Beyond nucleus, promptly for hole, tenuigenin and a matter, according to the predetermined color region color of classifying.For example, carry out the HSV conversion for R, G and the B value of each pixel, and these values are converted into the value of tone (H=0 to 1), saturation ratio (S=0 to 1) and lightness (V=0 to 1).Do not have nuclear color pixel and have lightness V＞0.92 and saturation ratio S＜0.2 if the pixel of being considered is classified as, then color of pixel is counted as the color (color_index=2) in hole.If pixel color is not nuclear color, color that neither pore, and have tone H＜0.9 and lightness V＜0.85, think that then this pixel color is cytoplasmic color (color_index=3).Under non-above-mentioned situation, confirm that this color of pixel is the color (color_index=4) of a matter.

Program 3

Candidate feature generation unit 1302 at first obtains s set of characteristic parameters, s=1 to N, and processing begins with s=1.If N_1+N_2+1≤s≤N, then candidate feature generation unit 1302 is confirmed the set of characteristic parameters of s to be provided unit 1303 to characteristic (color_index is th_s) with the identification number s (step S1402) of this candidate feature for x_s, y_s.

In this manual, (x_s, y_s) appointment is used for confirming the locations of pixels at the characteristic c of subimage, the color that color_index (=1 to 4) indication is corresponding with characteristic c, and th representes threshold parameter.

For the mode of learning (step S1403) that receives from mode of learning storage unit 1301, characteristic is confirmed the definite in the following manner characteristic c (step S1404) in unit 1303.Suppose and t mode of learning be represented as I_t (r i_rgb), then repeatedly carries out the operation that is used for confirming characteristic c for all modes of learning (M pattern).

Be used to confirm the color of pixel considered much at one, thereby omit its explanation corresponding to the processing of color_index (=1 to 4) and program 2.

At first, characteristic confirms that unit 1303 checks in the sub-image pixels position (x_s, near the pixel y_s); Promptly; (x_s, y_s) at interval whether the pixel in the scope of two pixels (x ', y ') has and color by the specified color-match of color_index with pixel on x direction of principal axis and y direction of principal axis.Pixel (x ', y ') for example with the pixel of being considered (x_s y_s) at interval in the scope of two pixels, and is represented as | x-x ' |≤2 draws | y-y ' |≤2.Characteristic is confirmed that unit 1303 is counted then near this pixel and is existed and have a color pixel with the color-match of color_index.Characteristic confirms that unit 1303 then with the quantity of the pixel sum divided by near the pixel this pixel, obtains MV thus.If MV surpasses threshold parameter (th_s), then characteristic c is set to 1.Otherwise characteristic c is set to 0.

Characteristic confirms that unit 1303 also can further use the characteristic of the pixel that following program confirms to be considered except three above-mentioned programs.

At first, characteristic confirms that 1303 pairs of unit comprise that R, G and the B value of each pixel in the subimage of the pixel of being considered carry out the HSV conversion, convert these values into the value of tone (H=0 to 1), saturation ratio (S=0 to 1) and lightness (V=0 to 1) then.Characteristic confirms that unit 1303 representes H, S and V value with for example 5 grades then.If each of these values of the pixel of for example being considered is equal to or less than 0.2 (=1/5), then color of pixel is expressed as (1,1,1).If H and S value be equal to or less than 0.2 and the V value be 0.2＜V≤0.4, then color of pixel is expressed as (1,1,2).When (x, when the pixel of y) locating had and be represented as the color of (H ', S ', V ') as set of characteristic parameters, characteristic c was " 1 " by the position in (x, y, H ', S ', V ') the specified subimage that receives from candidate feature generation unit 102.Otherwise characteristic c is " 0 ".

As above; When using s candidate feature (set of characteristic parameters) to each subimage calculated characteristics c; Characteristic confirms that unit 1303 calculates the mutual information amount MI that obtains from s candidate feature according to following expression (3), and the identification number s with candidate feature stores this amount MI (step S1405) then.

MI[Q；C]＝H[Q]-<H[Q|c]> _c

Wherein

H [Q] = - \underset{q}{Σ} P (q) \log P (q), P (q) = M (q) / M \cdot \cdot \cdot (3)

H [Q | c] = - \underset{q}{Σ} P (q | c) \log P (q | c),

P(q|c)＝M(q|c)/M(c)

In this, Q be classification set q=0, q=1}, and, the sum of M indication subimage.M (q) is the sum that belongs to the subimage of classification q, and M (c) indication is characterized in that the sum of the subimage of c, and (q c) is the sum that it is characterized in that c and belong to the subimage of classification q to M.

In expression formula (3), < H [Q|c]>c indication is for the equalization operation of c, and use following expression (4) is calculated it.

{< H [Q | c] >}_{c} = - \underset{c}{Σ} P (c) H [Q | c] . \cdot \cdot \cdot (4)

P(c)＝M(c)/M

, from candidate feature generation unit 1302 receive next candidate feature thereafter, that is, and (s+1) individual candidate feature, and processing (step S1402 to S1405) repeatedly in an identical manner.When accomplishing the calculating of mutual information amount MI for all candidate feature (N candidate), characteristic confirms that unit 1303 is compared to each other the mutual information amount MI of each candidate feature.Then, characteristic confirms that unit 1303 will have the peak Max MI [Q of mutual information amount; C] candidate feature confirm as first characteristic (step S1406) of the characteristic set that will confirm.

After first characteristic was confirmed, characteristic was confirmed unit 1303 definite second characteristics.In the same manner as described above, characteristic confirms that unit 1303 receives candidate feature (step S1402) from candidate feature generation unit 1302 in regular turn, and, come calculated characteristics c (step S1403 and S1404) according to each subimage.Perhaps, according to obtainable storage capacity, the calculation result of the characteristic c that can be stored in when confirming first characteristic at step S1404, and characteristic confirms that unit 1303 can read the data (candidate feature) of being stored.When for each subimage calculated characteristics c, characteristic confirms that unit 1303 uses s set of characteristic parameters at the known first characteristic c that has confirmed ₁Condition under calculate the mutual information amount MI that obtains from s candidate feature according to following expression (5) ₂, then with said amount MI ₂Identification number s with candidate feature stores (step S1405).

MI ₂[Q；C|C ₁]＝<H[Q|c ₁]> _c1

-<H[Q|(c，c ₁)]> _c1-c

Wherein

H [Q | c 1] = - \underset{q}{Σ} P (q | c_{1}) \log P (q | c_{1}), \cdot \cdot \cdot (5)

P(q|c ₁)＝M(q，c ₁)/M(c ₁)

H [Q | c, c 1] = - \underset{q}{Σ} P (q | c, c_{1}) \log P (q | c, c_{1}),

P(q|c，c ₁)＝M(q，c，c ₁)/M(c，c ₁)

In expression formula (5), M (c ₁) be that its first characteristic is c ₁The sum of subimage, M (q, c ₁) to indicate its first characteristic be c ₁And the sum that belongs to the subimage of classification q.M (c, c ₁) indication it is characterized in that c and its first characteristic are c ₁The sum of subimage.M (q, c, c ₁) be to it is characterized in that c, its first characteristic are c ₁And the sum that belongs to the subimage of classification q.

Through top operation, receive next candidate feature from candidate feature generation unit 1302, that is, and (s+1) individual candidate feature, and repeatedly handle this candidate feature (step S1402 to S1405) in an identical manner.When accomplishing the calculating of mutual information amount MI for all candidate feature (N candidate), characteristic confirms that unit 1303 will be from the condition mutual information amount MI of corresponding candidate feature acquisition ₂Make comparisons each other, and the peaked candidate feature that will have a mutual information amount is confirmed as the second characteristic c of the characteristic set that will confirm ₂(step S1406).

When confirming m characteristic in an identical manner, will maximize the valuation functions MI of following expression (6) _M+1Candidate feature as (m+1) individual characteristic c.

MI _m+1[Q；C|C ₁，C ₂，...，C _m]

＝<H[Q|c ₁，c ₂，...，c _m]> _{(c1，c2，...，cm)}…(6)

-<H[Q|(c，c ₁，c ₂，...，c _m)]> _{(c，c1，c2，...，cm)}

MI _M+1Be illustrated in known features (c ₁, c ₂..., c _m) condition under the quantity of information that obtains from characteristic c.Carry out continuously and should handle, up to the quantity of information that is obtained (other quantity of information) less than predetermined threshold MI_th, even select new characteristic.When threshold value MI_th for example is set to 0, repeatedly carry out above-mentioned handling procedure, to confirm next characteristic, become 0 up to the quantity of information that is obtained (other quantity of information), that is, and up to satisfying termination condition.

When satisfying termination condition, finish characteristic and confirm to handle.The parameter of each characteristic set of as above confirming is stored in (step S1407) in the characteristic storage unit 1304.

In the version that characteristic is confirmed to handle, can consider to reduce the program of the candidate feature quantity that produces by candidate feature generation unit 1302, be described below.For example, become the Gabor function again,, prepare the interior MV of class of the value of " a " that use expression formula (2) calculating in advance for the classification of q=0 and the classification of q=1 for each.Then, threshold value MI_th is set to the intermediate value between the MV in two classes.And for example, when when confirming first characteristic, using expression formula (3) for each compound Gabor function calculation mutual information amount MI, record provides the threshold value MI_th of maximum fault information MI for each candidate feature.When confirming second during with later characteristic, it is fixing that threshold value MI_th is held in processing.

Be used as the feature extraction function to obtain candidate feature though in this embodiment, will become the Gabor function again, can use another feature extraction function in addition.According to circumstances, only use such feature extraction function to obtain candidate feature.

As the favourable variation that characteristic confirm to be handled, can consider for example following variation: wherein, form the subspace, make the index that adds the distance that is used to be indicated to the subspace to candidate feature for each classification.And, also possibly be added near the weighted mean brightness of using the point that Gaussian function calculates to candidate feature.And; Might add normalized weighted mean brightness to candidate feature; Wherein the mean flow rate calculated of the Gaussian function through the use range broad comes normalization method to use near the weighted mean brightness of the point that Gaussian function calculates, and whether brighter or darker than its periphery this normalized weighted average brightness promptly be used to represent near this point index.And the standard feature of in diagnosis, using can be added on the candidate feature.

When the characteristic set of accomplishing the definite processing of characteristic and so confirm was stored in the characteristic storage unit 1304, becoming to produce sorted table 1306 (shown in Figure 16), to be used for pattern discrimination.Explain by the unit starting sorted table generation unit of expecting 1305 to set up the processing of sorted table 1306 now.

Sorted table generation unit 1305 at first receives each subimage and receives each parameters (below, suppose that the quantity sum of determined characteristic is n) of characteristic sets from characteristic storage unit 1304 from mode of learning storage unit 1301.Then, sorted table generation unit 1305 is stored the eigenwert (c of each subimage and each subimage in sorted table 1306 ₁, c ₂..., c _n).

Above program make the sorted table might produce each subimage that is used for classifying uniquely.More preferably, redundancy (not minding item) is used in expectation.For example, when use range only from the eigenwert (c of first eigenwert to an i eigenwert ₁, c ₂..., c _n) when classifying subimage, the value of the proper vector that (i+1) is individual and later replaces with the symbol that indication " is not minded " in sorted table.

Next referring to Figure 15, will explain through using redundancy (not minding) to produce the example of the program of sorted table 1306.Figure 15 illustrates the program of sorted table 1306 is set up in explanation in this embodiment the example of schema.

Sorted table generation unit 1305 at first uses the parameter of the characteristic set of storage in characteristic storage unit 1304 to come calculated characteristics vector (c for the subimage to its input ₁, c ₂..., c _n) (step S1501 and S1502).

Sorted table generation unit 1305 checks whether sorted table 1306 comprises the subimage (step S1503) with the proper vector of mating with top proper vector.When the territory of sorted table 1306 comprises the symbol that is used for indication " not minding ", then can regard any value corresponding as matching value with it.

As the result who confirms, if sorted table comprises the subimage of the proper vector of the proper vector coupling that has and calculated, then control turns back to step S1501, and in sorted table 1306, does not write down subimage to receive next subimage.

On the other hand, if there is not the subimage of coupling, then sorted table generation unit 1305 increases progressively variable i and is set to " 1 " (step S1504), to carry out following processing.At first, sorted table generation unit 1305 check classification under the subimage that belongs to and considered (for example, q=0) different classes (for example, q=1) subimage whether comprise comprise with said subimage all coupling first to i characteristic (c ₁, c ₂..., subimage ci) (step S1505).

As a result, if there is not such subimage, then first to i eigenwert (c ₁, c ₂..., c _i) (for example, q=0) be recorded in the sorted table together with the symbol of classification under the subimage.The symbol that is used for indication " not minding " is registered as the value (step S1506) of (i+1) and later proper vector.Control turns back to step S1501, to receive the subimage of next input.

On the other hand,, then add 1, and control turns back to step S1501 to increasing progressively variable i if there is such subimage.That is, carry out the increasing progressively of value of i continuously, become different through i eigenwert with other subimages up to the subimage of input.

Repeat top processing for all subimages.In this program, the situation of all subimages of impossible classification appears.For example, in some cases, the subimage that belongs to mutual different classes has equal proper vector.In this case, can carry out processing through following manner: for example, counting belongs to the quantity of the subimage of respective classes, and the classification under a large amount of subimages is confirmed as the classification that is associated with proper vector.

Also can adopt a kind of method, wherein, when increasing progressively the value of i, characteristic c ₁To c _iThe pattern that matches each other is classified as one group (group that comprises the subimage of lesser amt with acquisition in regular turn).When only comprising a number of sub images for one group, (i+1) individual and later characteristic of subimage can be set to not mind item.

Figure 16 illustrates the example of the sorted table 1306 that uses in the present invention.Figure 16 illustrates classification logotype mark (q) and the proper vector (c that is used to store each subimage ₁, c ₂..., c _n) form.In Figure 16, asterisk " * " indication " is not minded ".

Next referring to accompanying drawing, explanation is used the pathological image diagnostic method of sorted table.

Figure 17 is the block diagram that illustrates according to the treatment scheme of diagnostic method of the present invention.Figure 17 illustrates pattern input block 1701, feature extraction unit 1702 and diagnosis unit 1703.Figure 17 also illustrates: characteristic storage unit 1304, its storage will be used for being undertaken by feature extraction unit 1702 definite characteristic set of feature extraction; And, sorted table 1306, it is prepared in advance is used for diagnosis unit 1703 and diagnoses.

Pattern input block 1701 is the unit from the medium input subimage of expectation.In this embodiment, pattern input block 1701 is the unit of image (subimage) of cell that are used to import with the nucleus pathological tissue that is the center.Though input is the image at center with the nucleus in this embodiment, image is not limited thereto.For example, can be used as the subimage input and be used for carrying out the image that pathology is judged or confirmed in the diagnosis of pathological tissue, such as the image of nucleus, hole, tenuigenin and a matter etc. by the pathology expert.

Feature extraction unit 1702 is the subimages that send according to from pattern input block 1701, uses the unit of the characteristic of the said subimage of confirming of the incompatible extraction of feature set,

Diagnosis unit 1703 is the unit according to the information of being diagnosed said subimage to represent by the characteristic of feature extraction unit 1702 acquisitions.

At first, on the basis of the characteristic that obtains by pattern input block 1701, obtain subimage to be provided to feature extraction unit 1702.

Subsequently, the proper vector that feature extraction unit 1702 is calculated to the subimage of its input is provided to diagnosis unit 1703 with calculation result then.Use the characteristic set of confirming through program stored 1,2 in characteristic storage unit 1304 and 3, for example confirm the characteristic set that program is confirmed, come the calculated characteristics vector by above-mentioned characteristic.

Referring to sorted table 1306, the input item that diagnosis unit 1703 matees from its retrieval and proper vector, and read the class formative thing to export this mark as diagnostic result.If the input item that obtains from sorted table 1306 comprises " not minding ", then diagnosis unit 1703 is confirmed this this proper vector of input item coupling, and irrelevant with the value of this characteristic.

Confirm the characteristic of subimage and the advantage of the present invention of definite image in order further to illustrate through above-mentioned each program, with explanation in the present invention and the traditional method of using decision tree (ID3, the difference between C4.5).

The program of ID3 etc. and those similar being of the present invention: confirm the classifying rules in each node of decision tree according to the maximized standard of quantity of information.Yet, in ID3 and C4.5, confirm classifying rules (for example, characteristic) for each node.For example, when at definite first characteristic c ₁When second characteristic was confirmed in the back, classifying rules (characteristic) was being worked as c ₁Be 1 o'clock and work as c ₁It is the middle change of confirming between 0 o'clock.In contrast, according to the present invention,, then any n characteristic arbitrarily confirmed as and be equal to each other if use the node degree of depth that equates basically.This is the marked difference between the present invention and the traditional method.

In the present invention or in traditional method, mode of learning is by Complete Classification.Yet, in the generalize performance, promptly demonstrate marked difference on the performance of differentiation to the subimage of study not.Suppose that the tree degree of depth is substantially equal to one another (n) in two kinds of methods, when in ID3 or C4.5, confirming the individual characteristic of 2 < n >, only confirmed n characteristic in the present invention.That is, the present invention is structurally simple than traditional method.When problem becomes more complicated, being exponential form in the difference between the quantity of determined characteristic increases, and the required decision tree of program becomes darker.

If two sorters have the identical performance that is used for mode of learning, then have comparatively the generalize performance of the sorter of simple configuration and win, this is called as " Occam razor ".This is why to compare with traditional method, uses and confirms that according to the characteristic of the inventive method method and diagnostic method can improve the reason of generalize performance significantly.

Explanation now is used in mode of learning input block 1300 and the processing of pattern input block 1701 from pathological image extraction subimage.Though in this embodiment, having described with the nucleus is the extraction of the subimage at center, the invention is not restricted to this.Therefore, can extract morphological specificity part that the pathology expert pays close attention to as subimage in the pathological image of observing such as hole, tenuigenin and a matter.

Be used for extracting with the nucleus is that the processing of the subimage at center comprises: use the step of calculating miRNA dyeing signal in R, G and the B value of each pixel of pathological image; And, detect the step of nuclear central position according to the distribution of the miRNA of each pixel in this image dyeing signal.In fact, also comprise such as the processing that is used for level and smooth miRNA dyeing Signal Processing.Following embodiment relates to wherein, and nucleus is colored the situation into the pathological image of blueness through miRNA dyeing.

Next, referring to Figure 18, explanation is extracted the processing of subimage in mode of learning input block 1300 and pattern input block 1701.

When receiving pathological image as input (step S1801), mode of learning input block 1300 and pattern input block 1701 are colored to each pixel of the pathological image of blueness to nucleus wherein and distribute miRNA dyeing signal.

R, G and B value (R=0 to 255 when using 24 bits according to each pixel; G=0 to 255, B=0 to 255), calculate miRNA dyeing signal; This signal value in the zone of blue cell nuclear is 1.0, and value is 0 in other zones (dying the zone of different colours).In this is handled, check confirming by the nuclear distribution of color in the rgb space of R, G and B value representation, and calculate in the rgb value of each pixel and the distance between the distribution center.Specifically, check the rgb value of each pixel.If near the position in this value and the rgb space the nuclear distribution of color center is associated, then distribute the miRNA dyeing signal that has near 1 big value to this pixel.If should value in position away from the distribution of color center, then distribute the miRNA dyeing signal that has near 0 little value to this pixel.Yet, because nuclear coloration result is different between sample according to dying operation etc., so in adaptive approach, calculate nuclear distribution of color.

That is, through with reference to predetermined nuclear look zone, only wherein each pixel selection with the rgb value in nuclear look zone for being used to represent nuclear color pixel.

Confirm nuclear look zone in the following manner in advance.At first, collect and to close the nuclear image that colored state wherein changes owing to dying operation etc.Then, check rgb value for each pixel in the nuclear area of each image.Simultaneously, check to confirm for example in each image, to dye rgb value for the pixel in the zone of tenuigenin, the distinctive color of a matter and hole.Handle and to confirm not comprise or comprise dying for the zone of tenuigenin, the distinctive color of a matter and hole and comprise the nucleus look zone of the pixel of nuclear area seldom then.

Specifically, mode of learning input block 1300 distributes miRNA dyeing signal to each pixel in the following manner with pattern input block 1701.

At first, through with reference to predetermined nuclear look zone, be chosen in N the pixel (step S1802) that respectively has rgb value the nuclear look zone from pathological image to mode of learning

input block

1300 and 1701 inputs of pattern input block.Suppose that the rgb value of each of a selected N pixel comprises Ri, Gi and Bi (i=1 to N).Next, from Ri, Gi and the Bi (i=1 to N) of each pixel, according to following expression (7) calculate Ri, Gi and Bi MV (Ro, Go, Bo) with covariance matrix ∑ (step S1803).

R ₀＝1/N∑ _iR _i，G ₀＝1/N∑ _iG _i，B ₀＝1/N∑ _iB _i

∑＝1/N∑ _i(R _i-R ₀，G _i-G ₀，B _i-B ₀) ^T(R _i-R ₀，G _i-G ₀，B _i-B ₀)…(7)

In expression formula (7), T is the symbol that is used to indicate vectorial transposition.Use the covariance matrix ∑, (R, G is B) with MV (Ro, Go, the distance L between Bo) and the miRNA signal (Hema) (step S1804) that dyes to calculate each pixel according to following expression (8).

1/2(R-R ₀，G-G ₀，B-B ₀)∑ ^-1(R-R ₀，G-G ₀，B-B ₀) ^T

Hema＝exp(-1/2(R-R ₀，G-G ₀，B-B ₀)∑ ^-1(R-R ₀，G-G ₀，B-B ₀) ^T)…(8)

Next, explain that the distribution of using the miRNA dyeing signal that calculates for each pixel detects the processing of nuclear central position.

Use expression formula (8) to dye signal indication for the miRNA that each pixel obtains (r) as Hema.In this expression formula ,-r=(x, y) position vector of the location of pixels of indication in pathological image.According to using smoothing mask M _LowFollowing expression (9), for Hema (r) carry out smoothing operation (step S1805), deriving from the peak value of said smoothing, and this peak value be set to nuclear middle position (step S1806, S1807).

{Hema}^{'} (r) = Σ_{\overset{&RightArrow;}{r^{'}}} Hema (\overset{&RightArrow;}{r^{'}} - \overset{&RightArrow;}{r}) M_{low} (\overset{&RightArrow;}{r^{'}}) \cdot \cdot \cdot (9)

For example can use the function of expression formula (10) to realize smoothing mask M _Low

M_{low} (\overset{&RightArrow;}{r^{'}}) = M_{0} (\overset{&RightArrow;}{r^{'}}, s_{ex}) - M_{0} (\overset{&RightArrow;}{r^{'}}, s_{in}),

M_{0} (\overset{&RightArrow;}{r^{'}}, s) = 1 / l (When {| \overset{&RightArrow;}{r^{'}} |}^{2} \leq s^{2}), 0

(otherwise) ... (10)

Confirm the normalizing factor 1/l in expression formula (10) according to following expression (11).

Σ_{\overset{&RightArrow;}{r^{'}}} M_{0} (\overset{&RightArrow;}{r^{'}}, s) = 1 \cdot \cdot \cdot (11)

In this expression formula, S _ExAnd S _InIt is predetermined parameter.Usually, S _ExBe set to the value of about nuclear typical sizes (radius), and S _InBe set to S _ExAbout 1.2 times.

Calculate miRNA dyeing signal (miR ') for each pixel after; If the value of the miR ' of the point of being considered greater than predetermined threshold (for example; 0.25) and greater than near it (for example; In x and these three pixels of y wheelbase) the value of miR ' at any some place, then this point is detected to being set to the peak value (step S1807) of nuclear central point.

Consider the situation that has change in nuclear size, use to have different size (parameter S each other _ExAnd S _In) a plurality of (for example, three kinds) smoothing mask carry out miRNA dyeing signal smoothingization and detect with peak value.Peak by any detection of this operation can be set to nuclear center.

Mode of learning input block 1300 and pattern input block 1701 are through at first detecting nuclear center for carrying out above-mentioned processing to the pathological image of its

input.Unit

1300 and 1701 obtains the image (subimage) of a large amount of (quantity at the nucleus center that equals to be detected) of predetermined size from nucleus at the pathological image of picture centre then, is used as mode of learning or input pattern (step S1808) to extract each subimage.

Image diagnosis support system of the present invention also for example can comprise: the unit that is used to assess the validity of miRNA colored graph picture.Through the validity of assessment miRNA colored graph picture, can further improve the tolerance range of cancer assessment.The meaning of the assessment of the validity of the miRNA colored graph picture in this situation is identical with the assessment of the validity of carrying out the painted section slide glass of miRNA.Provide illustrated examples below.Yet, should be noted that to the invention is not restricted to this.

(the 3rd embodiment)

The image diagnosis support system that the 3rd embodiment relates to first or second embodiment further comprises: correcting unit, and it proofreaies and correct the colored state of miRNA colored graph picture; The non-tumor cell detecting unit, it detects the non-tumor cell in gauged miRNA colored graph picture; And, confirm the unit, it is confirmed the painted existence of miRNA in the non-tumor region that is detected or does not exist.

Can for example carry out the processing that the system by this embodiment carries out in the following manner.At first, proofread and correct the colored state of the miRNA colored graph picture that is obtained.For example, consider that the state of the state of employed section slide glass, painted in the same manner another section slide glass, the condition etc. of carrying out painted condition and obtaining view data proofread and correct for dyestuff or intensity etc.

Then, for gauged miRNA colored graph picture like this, carry out the detection of non-tumor cell.Can be for example based on discerning non-tumor cell such as the information of the shape of cell and size, nuclear shape and size and the position that in tissue, exists etc.Can carry out this identification through the module that has for example experienced machine learning based on above-mentioned condition.

Can carry out the detection of non-tumor cell through for example following manner: use counterstain to obtain the colored graph picture, and this is redyed image and miRNA dyeing images match.This coupling for example coupling with aforesaid miRNA colored graph picture and HE colored graph picture is identical.Can be according to for example coming suitably to confirm counterstain as the kind of the sample of subject.The example of counterstain comprises Kernechtrot.The kind of the non-tumor cell that detects is by special restriction, and, can be for example suitably confirmed according to kind as the sample of subject.The example of non-tumor cell comprises lymphocyte, inoblast and vascular endothelial cell.Whether satisfy specified cell size and/or shape according to these cells, might confirm whether subject is non-tumor cell.

Then, the miRNA dyeing of confirming detected non-tumor cell in miRNA colored graph picture exists or does not exist.As a result, when non-tumor cell receives miRNA dyeing, confirm that this miRNA colored graph picture is not effectively, and end treatment, and flow process is not proceeded to last step.On the other hand, when non-tumor cell does not receive miRNA when dyeing, confirm that this miRNA colored graph similarly is effectively, and this flow process proceeds to next procedure, for example, proceeding to dyes based on miRNA as stated detects the step of tumor region.

Unless otherwise indicated, can be with each above-mentioned embodiment combination with one another.

Embodiment

Below, through embodiment the present invention is described.Yet, should be noted that to the invention is not restricted to following embodiment.

[embodiment 1]

Use probe to carry out in situ hybridization, and, the expression level of the 2a of the has-miR-9 of inspection in acute myelocytic leukemia (AML) patient (n=4) and acute lymphoblastic leukemia (ALL) patient's (n=2) white corpuscle.

As probe, use the LNA modification utilize digoxin (DIG) mark probe (trade(brand)name: the miRCURY-LNA detection probes, Exiqon).Below, in the sequence that is used for the probe that has-miR-92a detects shown in the SEQ ID NO:5, and shown in the SEQ ID NO:6 as the sequence of the probe of negative control.The sequence of the negative control probe that illustrates below is the sequence that the sequence of the has-miR-92a detection probes that below upsetting, illustrates obtains.

Hsa-miR-92a detection probes (SEQ ID NO:5)

5’-acaggccgggacaagtgcaata-3’

Negative control probe (SEQ ID NO:6)

5’-gtgtaacacgtctatacgccca-3’

Use utilizes the Ventana Discovery robotization in situ hybridization appearance (trade(brand)name, Ventana Medical Systems) of RiboMap in situ hybridization test kit (trade(brand)name, Ventana Medical Systems) to carry out in situ hybridization.Unless otherwise indicated, carry out in situ hybridization according to the standard scheme that can provide from the RiboMap application specifications that Ventana Medical Systems (http://www.ventanamed.com) obtains.

At first, from each leukaemic's whole blood collection white cell.For the white corpuscle of collection like this, carry out the preparation of immobilization, paraffin embedding and the section of using the Paraformaldehyde 96 fixed solution according to the method for general use.Then,, section causes in situ hybridization after being taken off paraffin.In position in the hybridization,, carry out the initial fixationization that has experienced the section of taking off paraffin through using down RiboPrep (trade(brand)name, Ventana Medical Systems) based on Superlysoform to come incubation to have the slide glass 30 minutes of section at 37 ℃.Subsequently; Under 37 ℃ based on incubation slide glass in the RiboClear solution (trade(brand)name, Ventana Medical Systems) of hydrochloric acid 10 minutes, then; Promptly use proteolytic enzyme 2 (trade(brand)name, Ventana Medical Systems) to handle this slide glass 37 ℃ of following uses.Next, the prehybridization that slide glass is used for sex change under 70 ℃ was handled 6 minutes.Behind prehybridization, use hybridization buffer (trade(brand)name, Ventana Medical Systems) to handle slide glass 6 hours down at 37 ℃, so that cause the DIG mark of each slide glass 2ng ^*The hybridization of the probe of LNA modification.Then, slide glass uses down the cleaning 6 minutes that 2 * RiboWash solution (trade(brand)name, Ventana Medical Systems) hang down severity at 42 ℃.At 37 ℃ down use 1 * RiboFix (trade(brand)name, Ventana Medical Systems) clean slide glass 20 minute thereafter.Subsequently, under 37 ℃ with slide glass with the anti-DIG antibody of the biotin labeling of every slide glass 0.1 μ g (Sigma) incubation 30 minutes.Then, use streptavidin-SEAP combination (Dako) of every slide glass 0.1 μ g to come 37 ℃ of following incubation slide glasses 16 minutes., use BlueMap NBT/BCIP substrate reagent box (trade(brand)name, Ventana Medical Systems), carried out signal detection 4 hours down at 37 ℃ thereafter.At last, adjacent slide glass section is with Kernechtrot and HE counterstaining, and each slide glass is all covered by deckglass.

In its result shown in Figure 22.In Figure 22, the segment in up and middle row is the photo (FBA classification M3) that the leukocytic coloration result that obtains from AML patient is shown, and the segment in descending is the photo that the leukocytic coloration result that obtains from ALL patient is shown.Segment in left column is the photo that illustrates with Kernechtrot counterstaining result; Segment in middle row is that the photo that uses has-miR-92a detection probes coloration result is shown, and the segment that lists on the right side is that the photo that uses the negative control probe coloration result is shown, and the bar in each segment is that 50 μ m are long.

Can as can be seen from Figure 22 in AML patient and ALL patient, observe the dyeing of cell, and use the painted strength of signal of has-miR-92a detection probes greater than using the painted strength of signal of negative control probe.Though Figure 22 illustrates the result about single AML patient and single ALL patient acquisition, in remaining patient, obtains similar result.On the other hand, though not shown in the drawings, in normal white corpuscle, do not detect the expression of has-miR-92a.As stated, has-miR-92a strong expression in AML patient and ALL patient's white corpuscle.This has proved the possibility that can be evaluated at the cancer in the cell through the has-miR-92a of detection in white corpuscle.

Use the painted zone of has-miR-92a detection probes and use the painted zone of Kernechtrot identical.According to Kernechtrot dyeing, therefore the cancerous region that can dye, from this result, we can say, can be through detecting the possibility that has-miR-92a comes assessment of cancer.

[embodiment 2]

Collect the tissue that closes except using from breast, use with embodiment 1 in identical mode carry out the preparation of section and through in situ hybridization detection has-miR-92a.In its result shown in Figure 23.Figure 23 A to 23D is the photo that illustrates with respect to collect the miRNA coloration result of the different piece tissue that closes from breast.

In each of Figure 23 A to 23D, use arrow to indicate the part of coloured portions.Can find out from these figure, observe the dyeing of miRNA.

Though more than the present invention has been described with reference to illustrative embodiment and embodiment, the present invention never is limited to this.For for those skilled in the art, obviously can under the situation that does not depart from scope of the present invention, on configuration of the present invention and details, carry out various changes and modification.

The application requires the right of priority at the Japanese patent application No.2009-103332 of submission on April 21st, 2009.The whole disclosure of this Japanese patent application is comprised in this by reference.

[embodiment 3]

Collect the tissue that closes except using, use with embodiment 1 identical mode and carry out the preparation of section and detect has-miR-92a through in situ hybridization from liver cell.

Collect liver cell from hepatocellular carcinoma (HCC) patient (n=22) and non-tumour liver cirrhosis (LC) patient (n=5).Collect HCC patient's liver cell from the HCC patient of the kind with various ages, sex, hepatitis virus, clinical phase and tumour differentiation degree.

As a result, in non-tumour LC patient's liver cell, do not observe the dyeing of using the has-miR-92a detection probes basically.On the contrary, in all HCC patients' liver cell, observe the remarkable dyeing of using the has-miR-92a detection probes.

The liver cell that obtains from 22 HCC patients, Figure 24 illustrates about the painted result of the liver cell of two representative illustration (case 1 and case 2).Figure 24 illustrates the photo that has shown the hepatocellular coloration result that obtains from HCC patient.Segment in up and middle row illustrates the hepatocellular result about case 1, and the segment in descending illustrates the hepatocellular result about case 2.Segment in left column is the photo that illustrates with the result of Kernechtrot and HE counterstaining; Segment in middle row is that the photo that uses the painted result of has-miR-92a detection probes is shown, and the segment that lists on the right side is that the photo that uses the painted result of negative control probe is shown.Bar in each segment is that 100 μ m are long.Segment in middle row is the enlarged view of the segment in up.In each segment, the part that look dense is a coloured portions, and look light part is the part of being unstained.

Can as can be seen from Figure 24 in case 1 and case 2, all observe the dyeing of using the has-miR-92a detection probes.This proof has been expressed has-miR-92a in HCC patient's liver cell.Though Figure 24 only illustrates the result of two cases, the liver cell that obtains from all the other HCC patients, obtain similar result.

Use the painted zone of has-miR-92a detection probes and use Kernechtrot identical with the painted zone of HE.According to Kernechtrot dyeing, cancerous region can dye.Therefore, also, we can say from this result, can be through detecting the possibility that has-miR-92a comes assessment of cancer.

In addition, collect RNA from HCC patient's liver cell (n=5) and LC patient's liver cell (n=5), and measure the amount of expressed has-miR-92a through quantitative RT-PCR (qRT-PCR).And, as the endogenous contrast, measure the RNU48 amount of expression in an identical manner.Then, the ratio (hsa-miR-92a/RNU48) of the amount of the amount of expressed has-miR-92a and expressed RNU48 is calculated as the expression level of has-miR-92a.As a result, find that being expressed in from the liver cell that HCC patient obtains than the liver cell that obtains from LC patient of has-miR-92a is remarkable.

As stated, has-miR-92a strongly expressed in HCC patient's liver cell.This proof can be evaluated at the possibility of the cancer in the cell through the has-miR-92a of detection in liver cell.

Industrial applicability

According to the present invention,, become and for example to confirm existing or not existing of cancer generation or cancer progression with good reliability through in sample, detecting the expression level of cancer markers of the present invention.And through appraisal procedure of the present invention is associated with the for example cancer assessment through traditional HE dyeing etc., the cancer assessment of high reliability further becomes possibility.

Nomenclature

111: input unit

112: take-off equipment

113: the dyeing image data base

120: treatment unit

121: handling part is accepted in input

122: information acquisition portion

123: the images match handling part

124: tumor region extraction portion

125: stained positive cell containing ratio calculating part

130: storing device

131,132,133,134,135,136,1231: storage part

725: stained positive cell containing ratio and staining power calculating part

1126: tumour is confirmed and the tumor region calculating part

1213: the slide glass DB

1214: the slide glass device

1222: slide glass imaging portion

1300: the mode of learning input block

1301: the mode of learning storage unit

1302: the candidate feature generation unit

1303: characteristic is confirmed the unit

1304: the characteristic storage unit

1305: the sorted table generation unit

1306: sorted table

1701: the pattern input block

1702: feature extraction unit

1703: diagnosis unit

190: diagnostic imaging support equipment

191: handling part

192: storage part

193: microscope

194：CCD

195: scanning device

196: indicating meter

2001: image obtains the unit

2002: the information acquisition unit

2003: matching unit

2004: the tumor region designating unit

2005: computing unit

Claims

1. appraisal procedure that is used for the possibility of assessment of cancer, said appraisal procedure may further comprise the steps:

The cancer markers of detection in sample; And

Assess the possibility of the cancer in the sample based on the expression level of cancer markers, wherein

Said sample is a cell or tissue, and

Said cancer markers comprises at least a miRNA that is selected from has-miR-92 and has-miR-494.

2. appraisal procedure according to claim 1, wherein has-miR-92 is selected from hsa-miR-92a, hsa-miR-92a ^*, hsa-miR-92b and hsa-miR-92b ^*At least a.

3. appraisal procedure according to claim 2, wherein has-miR-92a is at least a among has-miR-92a-1 and the has-miR-92a-2.

4. appraisal procedure according to claim 2, wherein has-miR-92a ^*Be has-miR-92a-1 ^*And has-miR-92a-2 ^*In at least a.

5. appraisal procedure according to claim 1, wherein said miRNA is has-miR-92a.

6. according to each described appraisal procedure of claim 1 to 5, wherein cancer is to be selected from colorectal carcinoma, the rectum cancer, carcinoma of gallbladder, cancer of the stomach, mammary cancer, white blood disease, carcinoma of the pancreas, liver cancer, brain tumor and osteosarcomatous at least a cancer.

7. according to each described appraisal procedure of claim 1 to 6, wherein said appraisal procedure confirm the existence of cancer outbreak or do not exist, cancer progression phase or prognosis state.

8. according to each described appraisal procedure of claim 1 to 7, wherein detect in the step, through being selected from colour developing, fluorescence and autoradiographicly at least aly making said cancer markers visual in cancer markers.

9. according to each described appraisal procedure of claim 1 to 8, wherein sample is immobilized, and detects cancer markers through in-situ hybridization method.

10. according to each described appraisal procedure of claim 1 to 9, wherein represent the expression level of cancer markers by the amount of the cancer markers of expressing in the sample.

11. according to each described appraisal procedure of claim 1 to 10, wherein based on the expression level of the cancer markers that detects in the appraisal procedure, at least a method of following through being selected from (1), (2) and (3) is come the possibility of assessment of cancer:

(1) expression level with cancer markers in the expression level of cancer markers in person under inspection's sample and the normal person under inspection's sample compares; And when the expression level among the person under inspection is higher than the expression level among the normal person under inspection, confirm that the person under inspection has the high likelihood of cancer;

(2) expression level with cancer markers in the expression level of cancer markers in person under inspection's sample and the normal person under inspection's sample compares; And when the expression level among the person under inspection uprises than the expression level among the normal person under inspection relatively, confirm said person under inspection's relative late period of cancer; With

(3) expression level with cancer markers in each cancer patients's sample of the expression level of cancer markers in person under inspection's sample and different progressive stages compares; And, confirm that the cancer among the person under inspection is in identical progressive stage with cancer among the patient who shows identical or similar expression level.

12. according to each described appraisal procedure of claim 9 to 11, wherein

Detect in the step in cancer markers, obtain the painted cancer markers colored graph of cancer markers picture for said immobilization sample, and

Said appraisal procedure also comprises:

HE colored graph picture obtains step, and it obtains the HE colored graph picture for the immobilization sample;

The information acquisition step, it obtains the information of the tumor region in HE colored graph picture;

The coupling step, it calculates at the HE image and obtains HE colored graph picture that obtains in the step and the matched position that detects the cancer markers colored graph picture that obtains in the step in cancer markers;

Given step, it specifies in the tumor region in the cancer markers colored graph picture based on the information of the information of tumor region in the HE colored graph picture that obtains in the said information acquisition step and the matched position that in the coupling step, calculates; And

Stained positive cell detection step, the information that it is based on the tumor region in the specified cancer markers colored graph picture in the said given step detects the stained positive cell in the tumor region in cancer markers colored graph picture.

13. appraisal procedure according to claim 12; Wherein stained positive cell detection step is a calculation procedure; It is based in the given step information of tumor region in the specified cancer markers colored graph picture, calculates the stained positive cell containing ratio in the tumor region in cancer markers colored graph picture.